Re: intermittent amanda failure
Gene Heskett wrote: On Tuesday 09 March 2010, Dustin J. Mitchell wrote: On Wed, Mar 3, 2010 at 1:58 PM, Steve Wray steve.w...@cwa.co.nz wrote: Right, so the LATEST most up-to-date version of Debian uses a 3 year old version of amanda. Fantastic, thanks Debian for keeping things so 'stable'. To be fair, that's exactly the intent, and maintaining a Linux distribution is *not* easy. All of the binary-only distros are behind the times to varying degrees, although Debian is usually bringing up the rear of the bunch. I downloaded the actual latest stable version of amanda (2.6.1p2 from November 2009), compiled it and tested it. No bug. Yay! Thanks, Debian package maintainer. Not. [snip] If there are Amanda bugs that are holding back a version bump, please let me know. At the moment, I only see two open bugs, one from 2006 and one from 2008, neither of which is blocking a bump. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=500364 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=370319 Dustin I'm on your side here Dustin. The distros, debian in particular need prodding. What you use for a prod is up to you. :-) The problem I have with submitting bug reports to Debian on this sort of thing is this: If the bug is not security related then its extremely unlikely to be fixed until the NEXT stable release. In Debian, stability means making sure that non-security bugs are maintained throughout the lifetime of the release. The bugs are *part* of the stability. The theory is that people may have implemented workarounds for these bugs. If you go fixing the bug then you break their workaround. Since, due to this (and other ongoing concerns with the 'stability' problems of Debian), I will not be using the next stable release. So why would I bother to point out to them that, hey, maybe when you release the next Debian version you use the current version of Amanda? if I am not going to be using that version? This would be purely altruistic... and if Debian can't figure this out for themselves well... to be frank, I have no time for that. I go back 11+ years with amanda, usually running the bleeding edge as now. Considering that I build the new snapshot and use it nightly several times a week, the number of real bugs has been almost vanishingly small even when its labeled as alpha, not for production use. FWIW, 90% of those were tar's fault, not amanda's. There are several tar versions about, not all of which are even compatible with themselves. Amanda is compatible with itself with one exception, a format change a good 8 or 9 years ago. Folks like Dustin and Jean-Louis write tight, and correct code. I mentally salute them as I toss last nights printout on top of the stack (should, heaven forbid, I need to consult it) every morning. That reminds me; in one release of Debian the version of tar and of amanda were incompatible! It was the 'tar gives exit status 1 if a file changed while being read' problem IIRC. This was NEVER fixed in that 'stable' release. I should have seen the writing on the wall, really. -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
On Tuesday 09 March 2010, Steve Wray wrote: Gene Heskett wrote: On Tuesday 09 March 2010, Dustin J. Mitchell wrote: On Wed, Mar 3, 2010 at 1:58 PM, Steve Wray steve.w...@cwa.co.nz wrote: Right, so the LATEST most up-to-date version of Debian uses a 3 year old version of amanda. Fantastic, thanks Debian for keeping things so 'stable'. To be fair, that's exactly the intent, and maintaining a Linux distribution is *not* easy. All of the binary-only distros are behind the times to varying degrees, although Debian is usually bringing up the rear of the bunch. I downloaded the actual latest stable version of amanda (2.6.1p2 from November 2009), compiled it and tested it. No bug. Yay! Thanks, Debian package maintainer. Not. [snip] If there are Amanda bugs that are holding back a version bump, please let me know. At the moment, I only see two open bugs, one from 2006 and one from 2008, neither of which is blocking a bump. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=500364 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=370319 Dustin I'm on your side here Dustin. The distros, debian in particular need prodding. What you use for a prod is up to you. :-) The problem I have with submitting bug reports to Debian on this sort of thing is this: If the bug is not security related then its extremely unlikely to be fixed until the NEXT stable release. In Debian, stability means making sure that non-security bugs are maintained throughout the lifetime of the release. The bugs are *part* of the stability. What? Crazy theory, but probably correct, in the debian camp that is. The theory is that people may have implemented workarounds for these bugs. If you go fixing the bug then you break their workaround. Since, due to this (and other ongoing concerns with the 'stability' problems of Debian), I will not be using the next stable release. I probably will. For one thing, the 6.06 release and the accompanying releases of emc, which run my milling machine, haven't been getting all the loving that the 8.04 LTS version is getting, and at some point I'll have to update just so I can use the newer versions of EMC. OTOH, so far it is doing everything I ask it to do, soo. So why would I bother to point out to them that, hey, maybe when you release the next Debian version you use the current version of Amanda? if I am not going to be using that version? This would be purely altruistic... and if Debian can't figure this out for themselves well... to be frank, I have no time for that. Can't say as I blame you. Stone walls aren't much fun to talk to. I go back 11+ years with amanda, usually running the bleeding edge as now. Considering that I build the new snapshot and use it nightly several times a week, the number of real bugs has been almost vanishingly small even when its labeled as alpha, not for production use. FWIW, 90% of those were tar's fault, not amanda's. There are several tar versions about, not all of which are even compatible with themselves. Amanda is compatible with itself with one exception, a format change a good 8 or 9 years ago. Folks like Dustin and Jean-Louis write tight, and correct code. I mentally salute them as I toss last nights printout on top of the stack (should, heaven forbid, I need to consult it) every morning. That reminds me; in one release of Debian the version of tar and of amanda were incompatible! It was the 'tar gives exit status 1 if a file changed while being read' problem IIRC. This was NEVER fixed in that 'stable' release. I should have seen the writing on the wall, really. Yeah, we had to jump a tar version there, in addition to beating the tar folks about the brow, which I did at the time. They said they weren't gonna fix it, but then the crowd roar got to them and they did eventually. ;-) -- Cheers Dustin, Gene There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order. -Ed Howdershelt (Author) For most men life is a search for the proper manila envelope in which to get themselves filed. -- Clifton Fadiman
Re: intermittent amanda failure
On Wed, Mar 3, 2010 at 1:58 PM, Steve Wray steve.w...@cwa.co.nz wrote: Right, so the LATEST most up-to-date version of Debian uses a 3 year old version of amanda. Fantastic, thanks Debian for keeping things so 'stable'. To be fair, that's exactly the intent, and maintaining a Linux distribution is *not* easy. All of the binary-only distros are behind the times to varying degrees, although Debian is usually bringing up the rear of the bunch. I downloaded the actual latest stable version of amanda (2.6.1p2 from November 2009), compiled it and tested it. No bug. Yay! Thanks, Debian package maintainer. Not. Backup software is mission critical. Failing to track the upstream to this extent is simply unforgivable. I'm revising my opinion of Debian. I hear from a *lot* of folks on #amanda in exactly the same situation as you. Please do consider contacting the maintainer, or perhaps other Debian maintainers that might be able to poke the maintainer more effectively. I, as an upstream developer, don't have much impact on distro maintainers - apparently why don't you ship the latest release?! is a common refrain from upstreams. Distros aren't democracies, but they do listen to their users, and if enough people are asking why hasn't Amanda been bumped in 3 years? then someone with commit access will step up to take care of it. If there are Amanda bugs that are holding back a version bump, please let me know. At the moment, I only see two open bugs, one from 2006 and one from 2008, neither of which is blocking a bump. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=500364 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=370319 Dustin -- Open Source Storage Engineer http://www.zmanda.com
Re: intermittent amanda failure
On Tuesday 09 March 2010, Dustin J. Mitchell wrote: On Wed, Mar 3, 2010 at 1:58 PM, Steve Wray steve.w...@cwa.co.nz wrote: Right, so the LATEST most up-to-date version of Debian uses a 3 year old version of amanda. Fantastic, thanks Debian for keeping things so 'stable'. To be fair, that's exactly the intent, and maintaining a Linux distribution is *not* easy. All of the binary-only distros are behind the times to varying degrees, although Debian is usually bringing up the rear of the bunch. I downloaded the actual latest stable version of amanda (2.6.1p2 from November 2009), compiled it and tested it. No bug. Yay! Thanks, Debian package maintainer. Not. Backup software is mission critical. Failing to track the upstream to this extent is simply unforgivable. I'm revising my opinion of Debian. I hear from a *lot* of folks on #amanda in exactly the same situation as you. Amanda is not one of the distros favorite applications, having its own security model that has been at odds with the constraints of the packaging systems. Rpm in particular broke it regularly, and I long ago gave up helping the rpm folks who were determined to bend amanda to suit them. The tarball, OTOH, lends itself to the enterprising bash script writer, who can then install and test check the latest version of amanda in about 3, maybe 4 minutes on a fast machine using ccache. In fact I just installed the 20100308 snapshot of amanda-3.2alpha. And I used the same pair of scripts that has been installing amanda for me since about 2.5.1. Please do consider contacting the maintainer, or perhaps other Debian maintainers that might be able to poke the maintainer more effectively. I, as an upstream developer, don't have much impact on distro maintainers - apparently why don't you ship the latest release?! is a common refrain from upstreams. Distros aren't democracies, but they do listen to their users, and if enough people are asking why hasn't Amanda been bumped in 3 years? then someone with commit access will step up to take care of it. or you could build the tarball, which very nicely auto-configures amanda to run optimally on _your_ system. You have to setup a couple of files, maybe 3, and then just let the crontab of the amanda user take over from there. If configured to send the operator an email, you can 'read all about it' the next morning with your first cuppa. Whats not to like? My scripts are available, just ask. They are not 'big' scripts either. If there are Amanda bugs that are holding back a version bump, please let me know. At the moment, I only see two open bugs, one from 2006 and one from 2008, neither of which is blocking a bump. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=500364 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=370319 Dustin I'm on your side here Dustin. The distros, debian in particular need prodding. What you use for a prod is up to you. :-) I go back 11+ years with amanda, usually running the bleeding edge as now. Considering that I build the new snapshot and use it nightly several times a week, the number of real bugs has been almost vanishingly small even when its labeled as alpha, not for production use. FWIW, 90% of those were tar's fault, not amanda's. There are several tar versions about, not all of which are even compatible with themselves. Amanda is compatible with itself with one exception, a format change a good 8 or 9 years ago. Folks like Dustin and Jean-Louis write tight, and correct code. I mentally salute them as I toss last nights printout on top of the stack (should, heaven forbid, I need to consult it) every morning. -- Cheers, Gene There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order. -Ed Howdershelt (Author) Visits always give pleasure: if not on arrival, then on the departure. -- Edouard Le Berquier, Pensees des Autres
Re: intermittent amanda failure
I'd like to put in an update to this thread, for anyone interested. The backup server had been running Debian Etch. The version of amanda on Etch was giving the errors described in this thread. I upgraded the server to Debian Lenny. The problems still occured with the version in Lenny. Then I noticed that the version of amanda in Lenny dates from 2007. It gives its version as: 1:2.5.2p1-4 When I look at the amanda.org download page and compare version numbers I see this: 2.5.2p1 June 6 2007 Right, so the LATEST most up-to-date version of Debian uses a 3 year old version of amanda. Fantastic, thanks Debian for keeping things so 'stable'. I downloaded the actual latest stable version of amanda (2.6.1p2 from November 2009), compiled it and tested it. No bug. Thanks, Debian package maintainer. Not. Backup software is mission critical. Failing to track the upstream to this extent is simply unforgivable. I'm revising my opinion of Debian. Steve Wray wrote: Dustin J. Mitchell wrote: On Thu, Jan 21, 2010 at 5:05 PM, Jean-Louis Martineau martin...@zmanda.com wrote: xinetd is still configured to accept a tcp connection, but amandad expect a udp packet, so amandad do nothing and the server fail while waiting for an ACK. Right - it was the failure I expected to see, not Steve's bug. I'm not able to replicate Steve bug, many fix have been committed since 2.5.0p2. Good point. By my back-of-the-envelope calculations, we've fixed something like 500 bugs in the 4 years since 2.5.0 was released. Of course, we *introduced* some of of those bugs in those 4 years, too :) Steve: if you don't see something obvious that I missed in my replication effort, can you give this a try with a 2.6.1p2 server? I'm not able to make this test yet, perhaps later in the week. However, I've discovered that the amcheck problem may not actually reflect a genuine problem that affects backup. The other day, the cronjob amcheck reported these errors -- so for one thing its clearly intermittent. But the thing is that the backup job that ran that very night completed with no problems at all. -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
Dustin J. Mitchell wrote: On Thu, Jan 21, 2010 at 5:05 PM, Jean-Louis Martineau martin...@zmanda.com wrote: xinetd is still configured to accept a tcp connection, but amandad expect a udp packet, so amandad do nothing and the server fail while waiting for an ACK. Right - it was the failure I expected to see, not Steve's bug. I'm not able to replicate Steve bug, many fix have been committed since 2.5.0p2. Good point. By my back-of-the-envelope calculations, we've fixed something like 500 bugs in the 4 years since 2.5.0 was released. Of course, we *introduced* some of of those bugs in those 4 years, too :) Steve: if you don't see something obvious that I missed in my replication effort, can you give this a try with a 2.6.1p2 server? I'm not able to make this test yet, perhaps later in the week. However, I've discovered that the amcheck problem may not actually reflect a genuine problem that affects backup. The other day, the cronjob amcheck reported these errors -- so for one thing its clearly intermittent. But the thing is that the backup job that ran that very night completed with no problems at all. -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
On Wed, Jan 20, 2010 at 5:43 PM, Steve Wray steve.w...@cwa.co.nz wrote: The problem I had before was, to reiterate: Disklist with two entries. One entry uses bsdtcp The other entry uses BSD. If the client of the disklist entry that is configured on the server to use bsdtcp is not configured to use bsdtcp on the CLIENT in /etc/inetd.conf then the client which is configured to use default authentication returns errors. OK, this seems to be the simplest formation of this problem of errors with bsdtcp bleeding over into other, non-bsdtcp hosts. I tried to replicate it as follows. Disklist: euclid /etc { sysconfig-part auth bsdtcp } knuth /etc { sysconfig-part auth bsd } This works fine when both hosts are properly configured: Client check: 2 hosts checked in 1.130 seconds. 0 problems found. If I restart euclid's inetd config to run it with the wrong -auth parameter, then amcheck says: WARNING: euclid: selfcheck request failed: timeout waiting for ACK Client check: 2 hosts checked in 30.030 seconds. 1 problem found. so it doesn't give me any problems with knuth. If I stop xinetd on euclid (functionally equivalent to configuring it for bsd auth, since in bsd auth it would be listening on UDP port 10080, not TCP port 10080), the result is about the same: WARNING: euclid: selfcheck request failed: Connection refused Client check: 2 hosts checked in 10.047 seconds. 1 problem found. Am I missing something in the replication recipe? Dustin -- Open Source Storage Engineer http://www.zmanda.com
Re: intermittent amanda failure
Dustin J. Mitchell wrote: If I restart euclid's inetd config to run it with the wrong -auth parameter, then amcheck says: WARNING: euclid: selfcheck request failed: timeout waiting for ACK Client check: 2 hosts checked in 30.030 seconds. 1 problem found. xinetd is still configured to accept a tcp connection, but amandad expect a udp packet, so amandad do nothing and the server fail while waiting for an ACK. You must also change the following in xinetd: socket_type = dgram protocol= udp wait= yes I'm not able to replicate Steve bug, many fix have been committed since 2.5.0p2. Jean-Louis
Re: intermittent amanda failure
On Thu, Jan 21, 2010 at 5:05 PM, Jean-Louis Martineau martin...@zmanda.com wrote: xinetd is still configured to accept a tcp connection, but amandad expect a udp packet, so amandad do nothing and the server fail while waiting for an ACK. Right - it was the failure I expected to see, not Steve's bug. I'm not able to replicate Steve bug, many fix have been committed since 2.5.0p2. Good point. By my back-of-the-envelope calculations, we've fixed something like 500 bugs in the 4 years since 2.5.0 was released. Of course, we *introduced* some of of those bugs in those 4 years, too :) Steve: if you don't see something obvious that I missed in my replication effort, can you give this a try with a 2.6.1p2 server? Dustin -- Open Source Storage Engineer http://www.zmanda.com
Re: intermittent amanda failure
I am going to try and resurrect this thread having been able to home in on the apparent bug I found with bsdtcp auth. I have now converted most of our systems to using bsdtcp and amcheck is showing no errors at this time. The problem I had before was, to reiterate: Disklist with two entries. One entry uses bsdtcp The other entry uses BSD. If the client of the disklist entry that is configured on the server to use bsdtcp is not configured to use bsdtcp on the CLIENT in /etc/inetd.conf then the client which is configured to use default authentication returns errors. I now have a full disklist for all of our servers. Some are using bsdtcp others are using the default (which as I understand it is BSD). If I go to one of the clients whose disklist entry on the server is for bsdtcp and I change its /etc/inetd.conf from this: amanda stream tcp nowait backup /usr/lib/amanda/amandad amandad -auth=bsdtcp amdump to this: amanda dgram udp wait backup /usr/sbin/tcpd /usr/lib/amanda/amandad then ALL other disklist entries which are NOT using bsdtcp return errors on amcheck -c Disklist entries which are configured to use bsdtcp are not affected and do not generate errors, only those using BSD auth. This is just a change in inetd.conf on ONE client. The errors for the other disklist entries are: ERROR: an amanda client: [access as backup not allowed from root@the amanda server] amandahostsauth failed I have no idea if this is an amanda problem or a Debian packaging problem or what it is. But its repeatable. Version info for the amanda-server package on the amanda server is: r...@fileserver:~# dpkg -s amanda-server Package: amanda-server Status: install ok installed Priority: optional Section: utils Installed-Size: 1216 Maintainer: Bdale Garbee bd...@gag.com Architecture: i386 Source: amanda Version: 1:2.5.2p1-5 Steve Wray wrote: Steve Wray wrote: Jean-Louis Martineau wrote: Run 'amadmin CONFIG disklist' and check the auth is set as expected for all dles. I've done this, with the amanda.conf having bsdudp and with it having bsdtcp for that entry. In both cases all auth entries for all other DLE's are 'BSD'. In both cases only that one DLE is reported as having either bsdtcp or bsdudp, in both cases matching what is in the amanda.conf So I'd say that was all as expected. Having changed this one DLE to bsdudp I'm still seeing intermittent problems with the nightly backup run. I'd like to change to bsdtcp as it seems more robust but as I've mentioned there were some serious problems with that. Any further advice or help would be appreciated. Thanks. Jean-Louis Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) All dles for a client must have the same auth. different client can have different auth. We are going around in circles a little here. Allow me to try to make things very clear. In my amanda.conf I have a dumptype defined as such: (I've included the parent dumptypes. The 'global' dumptype is empty) define dumptype root-tar { global program GNUTAR comment root partitions dumped with tar compress none index exclude list /etc/amanda/exclude.gtar priority low } define dumptype nocomp-root-tar { root-tar comment Root partitions without compression compress none } define dumptype problem-nocomp-root-tar { nocomp-root-tar comment Root partitions without compression, problem client compress none auth bsdudp #auth bsdtcp } There are several DLEs for clients using the 'nocomp-root-tar' dumptype and only *one* DLE for *one* client using the 'problem-nocomp-root-tar' dumptype. With the bsdudp line uncommented everything is happy with an amcheck. With the bsdtcp line uncommented (and the bsdudp line commented out) *no* client is happy with the amcheck *other* than the client which uses 'problem-nocomp-root-tar'. However, as noted in another email this is intermittent, sometimes some clients using nocomp-root-tar are happy. So far its not exhibiting much pattern that I can see. The above *does* include the fact that I *do* change the inetd.conf on the client which uses problem-nocomp-root-tar *and* restart inetd. So, with a change of one line in a dumptype in a DLE used by one client, all other clients have problems.
Re: intermittent amanda failure
Steve Wray wrote: Jean-Louis Martineau wrote: Run 'amadmin CONFIG disklist' and check the auth is set as expected for all dles. I've done this, with the amanda.conf having bsdudp and with it having bsdtcp for that entry. In both cases all auth entries for all other DLE's are 'BSD'. In both cases only that one DLE is reported as having either bsdtcp or bsdudp, in both cases matching what is in the amanda.conf So I'd say that was all as expected. Having changed this one DLE to bsdudp I'm still seeing intermittent problems with the nightly backup run. I'd like to change to bsdtcp as it seems more robust but as I've mentioned there were some serious problems with that. Any further advice or help would be appreciated. Thanks. Jean-Louis Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) All dles for a client must have the same auth. different client can have different auth. We are going around in circles a little here. Allow me to try to make things very clear. In my amanda.conf I have a dumptype defined as such: (I've included the parent dumptypes. The 'global' dumptype is empty) define dumptype root-tar { global program GNUTAR comment root partitions dumped with tar compress none index exclude list /etc/amanda/exclude.gtar priority low } define dumptype nocomp-root-tar { root-tar comment Root partitions without compression compress none } define dumptype problem-nocomp-root-tar { nocomp-root-tar comment Root partitions without compression, problem client compress none auth bsdudp #auth bsdtcp } There are several DLEs for clients using the 'nocomp-root-tar' dumptype and only *one* DLE for *one* client using the 'problem-nocomp-root-tar' dumptype. With the bsdudp line uncommented everything is happy with an amcheck. With the bsdtcp line uncommented (and the bsdudp line commented out) *no* client is happy with the amcheck *other* than the client which uses 'problem-nocomp-root-tar'. However, as noted in another email this is intermittent, sometimes some clients using nocomp-root-tar are happy. So far its not exhibiting much pattern that I can see. The above *does* include the fact that I *do* change the inetd.conf on the client which uses problem-nocomp-root-tar *and* restart inetd. So, with a change of one line in a dumptype in a DLE used by one client, all other clients have problems. Perhaps I am misunderstanding something basic about dumptype configuration? -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
On Wed, Jan 06, 2010 at 04:15:41PM -0600, Dustin J. Mitchell wrote: On Wed, Jan 6, 2010 at 4:01 PM, Steve Wray steve.w...@cwa.co.nz wrote: Am I to understand that there could be a problem in having 'too many' DLE's for bsd or bsdudp to cope with? I never thought of there being a limit to the number of DLE's before... Our disklist file has 178. Yes, it's quite possible, and quite common with folks who have an Amanda built with only a small range of available ports. Caught my site, large number of DLE for a single client system, overall number of DLE across multiple cleints not an issue. With Dustin and Jean-Louis' direction I was able to config the single problem client to use TCP protocal and its been smooth sailing ever since. IMPORTANT NOTICE: This e-mail and any attachments may contain confidential or sensitive information which is, or may be, legally privileged or otherwise protected by law from further disclosure. It is intended only for the addressee. If you received this in error or from someone who was not authorized to send it to you, please do not distribute, copy or use it or any attachments. Please notify the sender immediately by reply e-mail and delete this from your system. Thank you for your cooperation.
Re: intermittent amanda failure
Steve Wray wrote: Dustin J. Mitchell wrote: I suspect an estimate or data timeout. Have you tried increasing dtimeout and etimeout? etimeout 2000 dtimeout 2000 I'd be surprised. These seem like fairly substantial values. 2000 seconds is roughly half an hour. I'll increase them by another 1000 seconds though, to see what happens. It failed again last night with 3000 second timeouts for these values. How can I verify whether theres a problem with timeouts? Thanks. -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Jean-Louis
Re: intermittent amanda failure
Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. I just tried following the instructions here: http://wiki.zmanda.com/index.php/How_To:Configure_bsd,_bsdtcp,_or_bsdudp_authentication I've added a single new dle config for this host: define dumptype problemserver-nocomp-root-tar { nocomp-root-tar comment Root partitions with compression compress none auth bsdtcp } and a single change to the disklist file to refer to this config entry for just that one single host. When I 'su - backup' and run 'amcheck -c' I get errors such as: ERROR: NAK an amanda client: user root from the amanda server is not allowed to execute the service noop: Please add amdump to the line in /var/backups/.amandahosts on the client for *every* amanda client in the list. It seems rather odd that this change for a single dle would introduce this error for every other dle? Its as if something has leaked out. I made no changes to any other configuration yet, just the change to the amanda server modifying a dle for a single client. I expected that it might fail for that client, but not for all of them... (I'll get to posting the debug logs soon). -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) Jean-Louis -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
On Wed, Jan 6, 2010 at 3:42 PM, Steve Wray steve.w...@cwa.co.nz wrote: Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) Well, all DLEs on a given host have to have the same auth. If you define different host-level parameters for DLEs on the same host, the results are undefined. The inetd configuration is on the client, and has to be set up to correspond with the auth you want to use. Check out Paul Yeatman's wonderful amanda-auth(7) for the nitty-gritty details. http://wiki.zmanda.com/man/amanda-auth.7.html Dustin -- Open Source Storage Engineer http://www.zmanda.com
Re: intermittent amanda failure
Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) All dles for a client must have the same auth. different client can have different auth. Jean-Louis
Re: intermittent amanda failure
Jean-Louis Martineau wrote: Steve Wray wrote: Dustin J. Mitchell wrote: On Wed, Jan 6, 2010 at 4:01 PM, Steve Wray steve.w...@cwa.co.nz wrote: Am I to understand that there could be a problem in having 'too many' DLE's for bsd or bsdudp to cope with? I never thought of there being a limit to the number of DLE's before... Our disklist file has 178. Yes, it's quite possible, and quite common with folks who have an Amanda built with only a small range of available ports. Ok I'm not sure about amanda as built for Debian but would it be safer for me to change pretty much our entire system to use bsdtcp? So far my experiments in this regard have not been great; adding just one single DLE using bsdtcp has caused all other DLE's to have problems. You can't change only one dle, you must change all dles for that client, you must also change inetd/xinetd on that client, don't change it on the server. This particular client has only one DLE. The other DLE's which are having problems are all for other clients. Ie: a change to bsdtcp in the DLE for one client has broken all other clients. -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
Dustin J. Mitchell wrote: On Wed, Jan 6, 2010 at 3:42 PM, Steve Wray steve.w...@cwa.co.nz wrote: Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) Well, all DLEs on a given host have to have the same auth. If you define different host-level parameters for DLEs on the same host, the results are undefined. On a given host, sure. In the case of bsdtcp it appears that there needs to be some config on the server to handle this (in inetd/xinetd on the server). I think thats what was causing the cascade of errors I posted before. The inetd configuration is on the client, and has to be set up to correspond with the auth you want to use. Check out Paul Yeatman's wonderful amanda-auth(7) for the nitty-gritty details. http://wiki.zmanda.com/man/amanda-auth.7.html Yes I've been reading that. This caught my eye: quote bsd communication and authentication The authentication is done using .amandahosts file in the Amanda user's home directory. The protocol between Amanda server and client is UDP. The number of disk list entries (DLEs)--number of Amanda clients--is limited by the UDP packet size. This authentication protocol will use a different port for each data stream (see PORT USAGE below) bsdudp communication and authentication The authentication is done using .amandahosts files in the Amanda user's home directory. It uses UDP protocol between Amanda server and client for data and hence the number of DLEs is limited by the UDP packet size. It uses one TCP port to establish the connection and multiplexes all data streams using one port on the server (see PORT USAGE below). /quote Am I to understand that there could be a problem in having 'too many' DLE's for bsd or bsdudp to cope with? I never thought of there being a limit to the number of DLE's before... Our disklist file has 178. -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
Steve Wray wrote: Dustin J. Mitchell wrote: On Wed, Jan 6, 2010 at 4:01 PM, Steve Wray steve.w...@cwa.co.nz wrote: Am I to understand that there could be a problem in having 'too many' DLE's for bsd or bsdudp to cope with? I never thought of there being a limit to the number of DLE's before... Our disklist file has 178. Yes, it's quite possible, and quite common with folks who have an Amanda built with only a small range of available ports. Ok I'm not sure about amanda as built for Debian but would it be safer for me to change pretty much our entire system to use bsdtcp? So far my experiments in this regard have not been great; adding just one single DLE using bsdtcp has caused all other DLE's to have problems. You can't change only one dle, you must change all dles for that client, you must also change inetd/xinetd on that client, don't change it on the server.
Re: intermittent amanda failure
On Wed, Jan 6, 2010 at 4:01 PM, Steve Wray steve.w...@cwa.co.nz wrote: Am I to understand that there could be a problem in having 'too many' DLE's for bsd or bsdudp to cope with? I never thought of there being a limit to the number of DLE's before... Our disklist file has 178. Yes, it's quite possible, and quite common with folks who have an Amanda built with only a small range of available ports. Dustin -- Open Source Storage Engineer http://www.zmanda.com
Re: intermittent amanda failure
Dustin J. Mitchell wrote: On Wed, Jan 6, 2010 at 4:01 PM, Steve Wray steve.w...@cwa.co.nz wrote: Am I to understand that there could be a problem in having 'too many' DLE's for bsd or bsdudp to cope with? I never thought of there being a limit to the number of DLE's before... Our disklist file has 178. Yes, it's quite possible, and quite common with folks who have an Amanda built with only a small range of available ports. Ok I'm not sure about amanda as built for Debian but would it be safer for me to change pretty much our entire system to use bsdtcp? So far my experiments in this regard have not been great; adding just one single DLE using bsdtcp has caused all other DLE's to have problems. And intermittent problems at that Sometimes 'amcheck -c' will report an error like: ERROR: NAK amanda client: user root from amanda server is not allowed to execute the service noop: Please add amdump to the line in /var/backups/.amandahosts on the client and the next time I run amcheck this error is not reported. Might be fine another run and then bad later. Whereas the amanda client which I configured for bsdtcp -- the only one I have thus configured so far -- does not return any error. (and yes, I am running amcheck from a 'su - backup' session where backup is the user that amanda runs as). -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
Jean-Louis Martineau wrote: Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) All dles for a client must have the same auth. different client can have different auth. We are going around in circles a little here. Allow me to try to make things very clear. In my amanda.conf I have a dumptype defined as such: (I've included the parent dumptypes. The 'global' dumptype is empty) define dumptype root-tar { global program GNUTAR comment root partitions dumped with tar compress none index exclude list /etc/amanda/exclude.gtar priority low } define dumptype nocomp-root-tar { root-tar comment Root partitions without compression compress none } define dumptype problem-nocomp-root-tar { nocomp-root-tar comment Root partitions without compression, problem client compress none auth bsdudp #auth bsdtcp } There are several DLEs for clients using the 'nocomp-root-tar' dumptype and only *one* DLE for *one* client using the 'problem-nocomp-root-tar' dumptype. With the bsdudp line uncommented everything is happy with an amcheck. With the bsdtcp line uncommented (and the bsdudp line commented out) *no* client is happy with the amcheck *other* than the client which uses 'problem-nocomp-root-tar'. However, as noted in another email this is intermittent, sometimes some clients using nocomp-root-tar are happy. So far its not exhibiting much pattern that I can see. The above *does* include the fact that I *do* change the inetd.conf on the client which uses problem-nocomp-root-tar *and* restart inetd. So, with a change of one line in a dumptype in a DLE used by one client, all other clients have problems. Perhaps I am misunderstanding something basic about dumptype configuration? -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
Run 'amadmin CONFIG disklist' and check the auth is set as expected for all dles. Jean-Louis Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) All dles for a client must have the same auth. different client can have different auth. We are going around in circles a little here. Allow me to try to make things very clear. In my amanda.conf I have a dumptype defined as such: (I've included the parent dumptypes. The 'global' dumptype is empty) define dumptype root-tar { global program GNUTAR comment root partitions dumped with tar compress none index exclude list /etc/amanda/exclude.gtar priority low } define dumptype nocomp-root-tar { root-tar comment Root partitions without compression compress none } define dumptype problem-nocomp-root-tar { nocomp-root-tar comment Root partitions without compression, problem client compress none auth bsdudp #auth bsdtcp } There are several DLEs for clients using the 'nocomp-root-tar' dumptype and only *one* DLE for *one* client using the 'problem-nocomp-root-tar' dumptype. With the bsdudp line uncommented everything is happy with an amcheck. With the bsdtcp line uncommented (and the bsdudp line commented out) *no* client is happy with the amcheck *other* than the client which uses 'problem-nocomp-root-tar'. However, as noted in another email this is intermittent, sometimes some clients using nocomp-root-tar are happy. So far its not exhibiting much pattern that I can see. The above *does* include the fact that I *do* change the inetd.conf on the client which uses problem-nocomp-root-tar *and* restart inetd. So, with a change of one line in a dumptype in a DLE used by one client, all other clients have problems. Perhaps I am misunderstanding something basic about dumptype configuration?
Re: intermittent amanda failure
Jean-Louis Martineau wrote: Run 'amadmin CONFIG disklist' and check the auth is set as expected for all dles. I've done this, with the amanda.conf having bsdudp and with it having bsdtcp for that entry. In both cases all auth entries for all other DLE's are 'BSD'. In both cases only that one DLE is reported as having either bsdtcp or bsdudp, in both cases matching what is in the amanda.conf So I'd say that was all as expected. Jean-Louis Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) All dles for a client must have the same auth. different client can have different auth. We are going around in circles a little here. Allow me to try to make things very clear. In my amanda.conf I have a dumptype defined as such: (I've included the parent dumptypes. The 'global' dumptype is empty) define dumptype root-tar { global program GNUTAR comment root partitions dumped with tar compress none index exclude list /etc/amanda/exclude.gtar priority low } define dumptype nocomp-root-tar { root-tar comment Root partitions without compression compress none } define dumptype problem-nocomp-root-tar { nocomp-root-tar comment Root partitions without compression, problem client compress none auth bsdudp #auth bsdtcp } There are several DLEs for clients using the 'nocomp-root-tar' dumptype and only *one* DLE for *one* client using the 'problem-nocomp-root-tar' dumptype. With the bsdudp line uncommented everything is happy with an amcheck. With the bsdtcp line uncommented (and the bsdudp line commented out) *no* client is happy with the amcheck *other* than the client which uses 'problem-nocomp-root-tar'. However, as noted in another email this is intermittent, sometimes some clients using nocomp-root-tar are happy. So far its not exhibiting much pattern that I can see. The above *does* include the fact that I *do* change the inetd.conf on the client which uses problem-nocomp-root-tar *and* restart inetd. So, with a change of one line in a dumptype in a DLE used by one client, all other clients have problems. Perhaps I am misunderstanding something basic about dumptype configuration? -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.
Re: intermittent amanda failure
I'll attach some debug logs. For the purposes of this test I cut the disklist file down to two entries: One entry is a client which is configured to use simple BSD auth. The other entry is a client which is configured to use bsdtcp auth. Both of these have been verified by running amcheck disklist. This file: amcheck.20100107131821.debug is from the client whose DLE has it using BSD auth. This file: amandad.20100107131826.debug is from the client whose DLE has it using bsdtcp auth. This file: amcheck.20100107131821.debug is from the server on that same run. The amcheck -c command reported: bac...@fileserver:~$ amcheck -c cwa-lto Amanda Backup Client Hosts Check ERROR: NAK zimbra1.internal.cwa.co.nz: user root from fileserver.internal.cwa.co.nz is not allowed to execute the service noop: Please add amdump to the line in /var/backups/.amandahosts on the client Client check: 2 hosts checked in 5.097 seconds, 1 problem found (brought to you by Amanda 2.5.2p1) Steve Wray wrote: Jean-Louis Martineau wrote: Run 'amadmin CONFIG disklist' and check the auth is set as expected for all dles. I've done this, with the amanda.conf having bsdudp and with it having bsdtcp for that entry. In both cases all auth entries for all other DLE's are 'BSD'. In both cases only that one DLE is reported as having either bsdtcp or bsdudp, in both cases matching what is in the amanda.conf So I'd say that was all as expected. Jean-Louis Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: Jean-Louis Martineau wrote: Steve Wray wrote: On the client, in the sendbackup.20100106012630.debug log I see: sendbackup-gnutar: time 0.056: /usr/lib/amanda/runtar: pid 3348 sendbackup: time 0.057: started backup sendbackup: time 90.352: index tee cannot write [Broken pipe] sendbackup: time 90.352: pid 3346 finish time Wed Jan 6 01:28:01 2010 90 seconds, it's not a dtimeout issue. Post all debug files for the run. You can also try the bsdtcp auth, it is more firewall friendly. Ah hang on, am I right in understanding that you can't have just one dle using bsdtcp auth? That they would all have to have it? (ie the inetd configuration) All dles for a client must have the same auth. different client can have different auth. We are going around in circles a little here. Allow me to try to make things very clear. In my amanda.conf I have a dumptype defined as such: (I've included the parent dumptypes. The 'global' dumptype is empty) define dumptype root-tar { global program GNUTAR comment root partitions dumped with tar compress none index exclude list /etc/amanda/exclude.gtar priority low } define dumptype nocomp-root-tar { root-tar comment Root partitions without compression compress none } define dumptype problem-nocomp-root-tar { nocomp-root-tar comment Root partitions without compression, problem client compress none auth bsdudp #auth bsdtcp } There are several DLEs for clients using the 'nocomp-root-tar' dumptype and only *one* DLE for *one* client using the 'problem-nocomp-root-tar' dumptype. With the bsdudp line uncommented everything is happy with an amcheck. With the bsdtcp line uncommented (and the bsdudp line commented out) *no* client is happy with the amcheck *other* than the client which uses 'problem-nocomp-root-tar'. However, as noted in another email this is intermittent, sometimes some clients using nocomp-root-tar are happy. So far its not exhibiting much pattern that I can see. The above *does* include the fact that I *do* change the inetd.conf on the client which uses problem-nocomp-root-tar *and* restart inetd. So, with a change of one line in a dumptype in a DLE used by one client, all other clients have problems. Perhaps I am misunderstanding something basic about dumptype configuration? -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this. amandad: debug 1 pid 5084 ruid 34 euid 34: start at Thu Jan 7 13:18:21 2010 Could not open conf file /etc/amanda/amanda-client.conf: No such file or directory amandad: time 0.000: security_getdriver(name=bsd) returns 0xb7f834c0 amandad: version 2.5.2p1 amandad: time 0.000: build: VERSION=Amanda-2.5.2p1 amandad: time 0.000:BUILT_DATE=Sat Aug 16 16:06:29 ART 2008 amandad: time 0.000:BUILT_MACH=Linux rover 2.6.25.15 #4 SMP Thu Aug 7 11:07:30 MDT 2008 i686 GNU/Linux amandad: time 0.000:CC=gcc amandad: time 0.000:CONFIGURE_COMMAND='./configure' '--prefix=/usr'
Re: intermittent amanda failure
I suspect an estimate or data timeout. Have you tried increasing dtimeout and etimeout? Dustin -- Open Source Storage Engineer http://www.zmanda.com
Re: intermittent amanda failure
Dustin J. Mitchell wrote: I suspect an estimate or data timeout. Have you tried increasing dtimeout and etimeout? etimeout 2000 dtimeout 2000 I'd be surprised. These seem like fairly substantial values. 2000 seconds is roughly half an hour. I'll increase them by another 1000 seconds though, to see what happens. -- Please remember that an email is just like a postcard; it is not confidential nor private nor secure and can be read by many other people than the intended recipient. A postcard can be read by anyone at the mail sorting office and expecting what is written on it to be private and secret is not realistic. Please hold no higher expectation of email. If you need to send confidential information in an email you need to use encryption. PGP is Pretty good for this.