subscribe amanda-users@amanda.org
[no subject]
We were running amanda 2.4.1p1 on solaris but we started getting timeout errors because we had more data than could be estimated(if I remeber correctly) within the timeout time. In amanda 2.4.1p1 this value could only be changed by recompiling amanda so to get round the problem we made the backup point futher down the tree so there was less data to be estimated in one go. For example in the disklist we changed /export/home/staff to /export/home/staff/finance, /export/home/staff/hr etc. This worked fine but we wanted to upgrade to version 2.4.1p2 of amanda which allowed more of the timeouts to be set in the amanda.conf. So at Xmas we upgraded to version 2.4.1p2 including updating amanda.conf with with the new values for timeouts etc. All seemed to to go well(and everything up until now has just been background info) as the report produced by amanda at the end of dumps did not produce errors. The trouble started when we needed to do a restore(I know I should have done a test restore in the first place) and we found that the tape had the directory structure i.e it had /export/home/staff/hr/bob but it did not have any of bob's files. After checking the end of dump reports again the size of /export/home/staff was very small i.e. maybe enough just to store the directory structure. Other parts of /export/home (/export/home is all stored on one disk) backed up fine. The total size of /export/home/staff is 8.9GB but the end of dump reports are not giving any errors about timeouts so why is amanda only backing up the directory structure and not the files, also why is it only happening for staff. Thanks in advance - David Flood Systems Administrator [EMAIL PROTECTED] Tel: +44 (0)1224 262721 Robert Gordon University School of Computing St. Andrews Street Aberdeen -
[no subject]
auth 064285c5 subscribe amanda-users [EMAIL PROTECTED] - David Flood Systems Administrator [EMAIL PROTECTED] Tel: +44 (0)1224 262721 Robert Gordon University School of Computing St. Andrews Street Aberdeen -
nmap
Should an AMANDA client have port 10080 open? When I run nmap on my AMANDA server (which is also an AMANDA client) I only see ports 10082 amandaidx and 10083 amidxtape. Port 10080 is shown to be closed. Should 10080 be open? Thanks!
Re: amcheck FIXED!!!
On Thursday 07 February 2002 15:19, R. Bradley Tilley wrote: When I run su amanda -c amcheck daily the backup server times out with the selfcheck request timed out error. Believe it or not, this just started happening, and I _haven't_ changed anything. The funny part is this: amdump works fine and runs nightly! It backs-up the server, even tho amcheck says it's down. I'm using the latest stable version with RH7.2 The system has been running fine for the last month (it's only a test setup so this isn't that big of a problem). I have rechecked .amandahosts; made sure xinetd was setup properly; emptied the hosts.deny file. Added ALL: ALL to the hosts.allow file, I doubled the etimeout from 300 to 600, but amcheck still times out. Has anyone else ever encountered a similar problem? Thanks, Brad It was RH's fault!. I did an up2date on the rpm packages, and one of the packages overwrote my hosts file in /etc. When I recreated the hosts file as it was before the package upgrade, amanda worked fine! Lesson: Beware of updates!!! Do them, but be prepared to troubleshoot your apps because upgrades _can_ break things.
amdump question
Hi, I have problem with amdump: host which I try to dump does not respond (by time-out) but it is alive. aaa | 309 less log.20020208.2 START planner date 20020208 INFO planner Adding new disk dmps.ripn.net:/dev/rsd0a. START driver date 20020208 ERROR taper no-tape [cannot overwrite active tape Daily00] FAIL planner local /dev/rsd0a 0 [Request to local timed out.] ^ FINISH planner date 20020208 WARNING driver WARNING: got empty schedule from planner STATS driver startup time 30.004 Would you be so kind to promt me possible reasons? Sincerely, Mary
Re: nmap
Should an AMANDA client have port 10080 open? Only if you want it to work :-). When I run nmap on my AMANDA server (which is also an AMANDA client) I only see ports 10082 amandaidx and 10083 amidxtape. Port 10080 is shown to be closed. ... TCP or UDP? 10082 and 10083 are TCP and should be ready to accept connections on a server (but not on a client). 10080 is UDP and should be ready to accept connections on a client. $ config.guess sparc-sun-solaris2.6 $ netstat -an | egrep '^TCP|^UDP|\.1008[0-9]' UDP *.10080 Idle TCP *.10082 *.*0 0 0 0 LISTEN *.10083 *.*0 0 0 0 LISTEN This assumes you have not changed the suggested service name or port numbers, etc. John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: [data timeout]
On Thu, Feb 07, 2002 at 12:01:03PM -0500, Joshua Baker-LePain wrote: On Thu, 7 Feb 2002 at 11:47am, Benjamin Gross wrote I've been having problems getting a successful level 0 backup from one of our servers. The amreport shows the following: FAILURE AND STRANGE DUMP SUMMARY: lopt /dev/hda1 lev 0 FAILED [data timeout] lopt /dev/hda1 lev 0 FAILED [dump to tape failed] Here are some of the FAILED AND STRANGE DUMP DETAILS: DUMP: Date of this level 0 dump: Wed Feb 6 23:22:40 2002 ... DUMP:Volume 1 started with block 1 at: Wed Feb 6 23:25:16 2002 ... DUMP 2.42% done at 1455 kB/s, finished in 3:21 ... DUMP: 61.83% done at 1543 kB/s, finished in 1:14 This is were details ends. Incidently, it is about this far into the dump, when dump fails. just a post in case anyone else experiences this problem. dump started to fail on this host (mandrake linux host) after an upgrade was performed on it. the kernel remained 2.2, but most of the binaries were upgraded. in this case dump version 0.4b23 (file size 55196) was installed. after i downgraded dump to version 0.4b19 (file size 48028 - version that was on the host before the upgrade), dump worked successfully. thanks for everyone's help. ben
Re: Some stuff not being backed up
This worked fine but we wanted to upgrade to version 2.4.1p2 ... Do you mean 2.4.2p2? That's the latest stable release. All seemed to to go well(and everything up until now has just been background info) as the report produced by amanda at the end of dumps did not produce errors. The trouble started when we needed to do a restore ... Isn't that just always the case :-) :-). ... we found that the tape had the directory structure i.e it had /export/home/staff/hr/bob but it did not have any of bob's files. Do you mean you got back (multiple) subdirectories underneath /export/home/staff/hr/bob, but just no files in those directories? Or do you mean you had the top level .../bob but nothing at all under that? When you say the tape had ..., do you mean you were using amrecover or that you actually did a restore from the tape? After checking the end of dump reports again the size of /export/home/staff was very small ... Looking at your disklist, find the client backing up /export/home/staff. Going to that client, do a df of /export/home/staff. Is /export/home/staff a real disk on that client, or is it NFS mounted onto it? What are you using to back up /export/home/staff, dump or GNU tar? What kind of OS is running on the client? What version of dump or GNU tar are you using? Look at one of the /tmp/amanda/sendbackup*debug files for the backup of this file system and see which command (and args) were used and whether it reported any problems. (/export/home is all stored on one disk) ... So /export/home/staff is just a subdirectory of /export/home? It's not a mount point for some other disk? What's in the disklist, /export/home or /export/home/staff? Where I'm headed with much of this is that backup programs (the way Amanda runs them) do *not* cross mount points. Also, dump will not back up NFS areas at all, and GNU tar may have trouble with NFS because it runs as root. why is amanda only backing up the directory structure and not the files ... Keep in mind it's not Amanda doing the backup. Amanda is running some other program that's doing the backups. David Flood John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: [data timeout]
Thanks for the followup. ... in this case dump version 0.4b23 (file size 55196) was installed. after i downgraded dump to version 0.4b19 (file size 48028 - version that was on the host before the upgrade), dump worked successfully. It's my understanding dump is an active project. You should report this to those folks so they can work on it. If they don't know about it, it will never get fixed :-). ben John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Everything seems right, except...
Hi, all. The files permissions' are right to me. The .amandahosts are correct in server and client. When cron is execute amdump, fabio (not a backup user) user received from root a e-mail like this: Subject: Cron amanda@sam /usr/sbin/amdump DailySet1 amanda.conf, line 26: configuration keyword expected amanda.conf, line 26: end of line expected amanda.conf, line 28: configuration keyword expected amanda.conf, line 28: end of line expected amanda.conf, line 31: configuration keyword expected amanda.conf, line 31: end of line expected amanda.conf: Bad file descriptor amanda.conf, line 26: configuration keyword expected amanda.conf, line 26: end of line expected amanda.conf, line 28: configuration keyword expected amanda.conf, line 28: end of line expected amanda.conf, line 31: configuration keyword expected amanda.conf, line 31: end of line expected amanda.conf: Bad file descriptor amanda.conf, line 26: configuration keyword expected amanda.conf, line 26: end of line expected amanda.conf, line 28: configuration keyword expected amanda.conf, line 28: end of line expected amanda.conf, line 31: configuration keyword expected amanda.conf, line 31: end of line expected amanda.conf: Bad file descriptor /usr/sbin/amdump: [: amanda: unary operator expected /usr/sbin/amdump: /amdump: Permission denied This lines are: 25: etimeout 300 26: dtimeout 1800 27: ctimeout 50 28: 29: tapebufs 20 30: runtapes 1 # explained in WHATS.NEW 31: #tpchanger no-changer # the tape-changer glue script, see TAPE.CHANGERS And my amanda.conf is in /etc/amanda. Thanks for any help!! Fabio.
Re: amanda client service
i work at a sercurity company. and was told to configure a backup for the company's data. Now *that* sounds like a good plan :-). When I start amcheck on the server now, the service amanda on the clients gets shut down with following error: Feb 8 13:25:04 client [767]: amanda service was deactivated because of looping Feb 8 13:25:04 client [767]: recv: Bad file descriptor (errno = 9) I *think* the second message (recv) is just a side effect of the first problem, but you might want to report it to the xinetd folks anyway. That's coming from them, not Amanda. What does you Amanda xinetd.conf file look like? Does /tmp/amanda/amandad*debug get created on the client? What happens if you run this on the client: su AMANDA -c /PATH/TO/amandad It should just sit there for 30 seconds and then terminate. It should also create a /tmp/amanda/amandad*debug file that ends with a timeout error (which is normal for this test). In your archive i found that I should add this entry to my /etc/rc.d/init.d/xinetd: EXTRAOPTIONS=-loop 100 I don't think that would help. It will just make xinetd keep trying, but the real problem is that amandad is terminating. The amcheck out produces a second error. I do not know whether this error depends on the shut down of the client: ... ERROR: localhost NAK: amandad busy WARNING: IP_OF_MY_TESTCLIENT: selfcheck request timed out. Host down? You said you are just testing with the client and the server. I take it those are two different machines? And do you mean that you have entries for both localhost and the client in your disklist? One minor point first. You should not use localhost in disklist. Use the real host name of the server. Amanda remembers backups by their host name and disk. If you need to move things around at some point, localhost may point to some other machine. The NAK error means an amandad was still running on localhost when you tried to run amcheck. I assume you didn't try to run two amcheck's at the same time? Look for any stray amandad processes running on localhost and kill them. Then try the amcheck again. Leonie P.S: Sorry, I know, my English is ugly ((-: Your English is just fine. You should see some of the things my colleagues write who have lived here their whole life :-). i.A. Leonie Nuenighoff John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Q. R.e. smb amanda
So, I've got a share called web sitting on a box. Inside web are 28gb. Eventually, I'll back up the whole thing, but at the moment, I'd like to only backup a directory called cos. If I specify cos, it seems to bomb: In my disklist: adv2.kellogg.nwu.edu//adv.kellogg.nwu.edu/web/cos nocomp-user-gnutar In my /etc/amandapass: //adv.kellogg.nwu.edu/web/cos amanda%xx But... Amanda Backup Client Hosts Check ERROR: adv2.kellogg.nwu.edu: [cannot find password for //adv.kellogg.nwu.edu/web/cos] Client check: 1 host checked in 0.950 seconds, 1 problem found (brought to you by Amanda 2.4.3b2) Why it doesnt find it is beyond me. The sharename is /web - I'm guessing this bombs because I cant specify a directory structure after the /web ? If I remove /cos from both disklist and amandapass and have this: Disklist: adv2.kellogg.nwu.edu//adv.kellogg.nwu.edu/web nocomp-user-gnutar Amandapass: adv.kellogg.nwu.edu/web/ amanda%xx It SEEMS ok, but not really... Amanda Backup Client Hosts Check Client check: 1 host checked in 1.211 seconds, 0 problems found (brought to you by Amanda 2.4.3b2) But here's the other thing - it says 1 host checked, but I've got TWO hosts under my disklist ... so whats up with that... Whatever. So, I try running a dump anyway, and i get this: - this is a tapeless setup btw, if that makes any difference. FAILURE AND STRANGE DUMP SUMMARY: adv2.kello //adv.kellogg.nwu.edu/web lev 0 FAILED [can't switch to incremental dump] The other samba share backs up fine... HOSTNAME DISKL ORIG-KB OUT-KB COMP% MMM:SS KB/s MMM:SS KB/s -- - adv2.kellogg -delormdocs 11690 1690 --0:52 32.2 N/A N/A adv2.kellogg -wu.edu/web 0 FAILED --- For the record, adv is a unix box running samba. I know I can install the client but for right now I'm just trying to fool around and get a feel for how things work... Ideas?
follow up
ok so the problem is that it thinks I have no new tapes, I'd like to just write over tape 1 - how to I tell it to just write over it?
Missing Tape has Backups Screwed
One of my tapes did not come back from offsite in time for today's backup (Some winidiot sent if off for four weeks instead of one) so today's backup failed. How can I get amanda to just skip that tape and go on to the next one in the sequence? When I try amcheck, amanda thinks all of the tapes in the changer are active tapes and will not write to them. Even if I load the next tape in the sequence mnaually, amcheck rejects it. How can I tell which tapes it will write too? dumpcycle 7 days runspercycle 5 tapecycle 15 tapes runtapes 2 So far, backups have only needed one tape per run so it seems to me that tapes from two weeks ago should no longer be active but amanda thinks they are. -- Stephen Carville UNIX and Network Administrator DPSI (formerly Ace USA Flood Services) 310-342-3602 [EMAIL PROTECTED]
RE: Missing Tape has Backups Screwed
amadmin YOURCONFIG no-reuse MISSINGTAPE Then use the next tape in sequence. Just remember to mark it 'reuse' before it is next scheduled. -Original Message- From: Stephen Carville [mailto:[EMAIL PROTECTED]] Sent: Friday, February 08, 2002 2:37 PM To: Amanda Users Subject: Missing Tape has Backups Screwed One of my tapes did not come back from offsite in time for today's backup (Some winidiot sent if off for four weeks instead of one) so today's backup failed. How can I get amanda to just skip that tape and go on to the next one in the sequence? When I try amcheck, amanda thinks all of the tapes in the changer are active tapes and will not write to them. Even if I load the next tape in the sequence mnaually, amcheck rejects it. How can I tell which tapes it will write too? dumpcycle 7 days runspercycle 5 tapecycle 15 tapes runtapes 2 So far, backups have only needed one tape per run so it seems to me that tapes from two weeks ago should no longer be active but amanda thinks they are. -- Stephen Carville UNIX and Network Administrator DPSI (formerly Ace USA Flood Services) 310-342-3602 [EMAIL PROTECTED]
Re: follow up
I believe that you will need to do a amrmtape the tape from the database trhen you could do a amlabel and re-label the disk.and then you will need to reintroduce it back into the rotation Don [EMAIL PROTECTED] wrote: ok so the problem is that it thinks I have no new tapes, I'd like to just write over tape 1 - how to I tell it to just write over it?
Re: follow up
What I'm looking to do is something somwhat automatic - so I don't have to do this every four days. I.e. is there a way to tell amdump to overwrite a backup tape? I.e. something like amdump -force DailySet1 If so, then I wouldn't have to do all this amlabel crap. Also, is there a way to get amanda to make .tar.gz's instead of its own format? In addition, is there a way tt amanda to run several amandas at once? I'm backing up stuff thats on a slow connection, so I've got plenty of bandwith.. Sine amanda does stuff sequentially, I'm stuck waiting 20 minutes for some file to backup and im only using like uhm, 20k/s... So I could be running 10 other amanda backups at teh same time Do I just make multiple confs and put them all in my crontabs? cosimo On Fri, 8 Feb 2002, Don Potter wrote: I believe that you will need to do a amrmtape the tape from the database trhen you could do a amlabel and re-label the disk.and then you will need to reintroduce it back into the rotation Don [EMAIL PROTECTED] wrote: ok so the problem is that it thinks I have no new tapes, I'd like to just write over tape 1 - how to I tell it to just write over it?
Re: Everything seems right, except...
When cron is execute amdump, fabio (not a backup user) user received from root a e-mail like this: Subject: Cron amanda@sam /usr/sbin/amdump DailySet1 There are two many variables in the above :-). The Subject line implies the cron job is being run by amanda. The mail came from root (which might just be the way cron works on your system). Yet fabio got the mail. Can you determine which user actually tried to run /usr/sbin/amdump DailySet1? amanda.conf, line 26: configuration keyword expected amanda.conf, line 26: end of line expected ... What happens if you run /usr/sbin/amgetconf DailySet1 logdir as the same user that ran the cron job (I assume amgetconf is in the same directory as amdump)? Look at the beginning of /usr/sbin/amdump (it's a shell script) for the assignment to variable confdir. What is it set to? Is it set to /etc/amanda? Do /usr/sbin/amdump and /usr/sbin/amgetconf have roughly the same modification timestamp? If so, what is the SUF variable set to in amdump (just a few lines after the confdir assignment)? Fabio. John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: follow up
I think you can simply tell amanda to skip using that tape for the time being. /usr/local/sbin/amcheck config no-reuse tape-label and restore it to rotation when you are ready /usr/local/sbin/amcheck config reuse tape-label Note, tapes are used from eldest (bottom of the list) /usr/local/etc/amanda/config-name so your original order will not be preserved... I believe that you will need to do a amrmtape the tape from the database trhen you could do a amlabel and re-label the disk.and then you will need to reintroduce it back into the rotation Don [EMAIL PROTECTED] wrote: ok so the problem is that it thinks I have no new tapes, I'd like to just write over tape 1 - how to I tell it to just write over it?
Re: Can you change /tmp/amanda to a more permanent directory?
At 04:38 AM 2/7/2002 -0500, Joshua Baker-LePain wrote: On Thu, 7 Feb 2002 at 10:19am, Sascha Wuestemann wrote I found out, that a reboot of the amandaserver causes the night following backup to fail, because /tmp is emptied by the system after wakeup. . --with-tmpdir[=/temp/dir] area Amanda can use for temp files If a system is erasing /tmp (on a reboot, or on any schedule) does Amanda recreate its /tmp/amanda directory, or expect it to already be there? Is that where the problem arises? (still reading re-reading; wondering if this is going to cause me probs too when I try my first configurations) Deb Baddorf --- Deb Baddorf [EMAIL PROTECTED] 840-2289 You can't help getting older, but you don't have to get old. - George Burns IXOYE
RE: Missing Tape has Backups Screwed
On Fri, 8 Feb 2002, Bort, Paul wrote: - amadmin YOURCONFIG no-reuse MISSINGTAPE - - Then use the next tape in sequence. - - Just remember to mark it 'reuse' before it is next scheduled. Didn't work. $ amadmin daily1 tape The next Amanda run should go onto tape C83173 or a new tape. The next Amanda run should go onto tape C83174 or a new tape. $ amadmin daily1 no-reuse C83173 amadmin: marking tape C83173 as not reusable. $ amtape daily1 label C83174 amtape: scanning for tape with label C83174 amtape: slot 9: date 20020121 label C83174 (exact label match) amtape: label C83174 is now loaded. amdump reports *** A TAPE ERROR OCCURRED: [new tape not found in rack]. When I try 'amcheck daily1' it reports that _all_ the tapes are active. Even tapes that 'amtape daily1 info' says are not part of the current backup. -- Stephen Carville UNIX and Network Administrator DPSI (formerly Ace USA Flood Services) 310-342-3602 [EMAIL PROTECTED]
Re: Missing Tape has Backups Screwed
Stephen Carville wrote: One of my tapes did not come back from offsite in time for today's backup (Some winidiot sent if off for four weeks instead of one) so today's backup failed. How can I get amanda to just skip that tape and go on to the next one in the sequence? When I try amcheck, amanda thinks all of the tapes in the changer are active tapes and will not write to them. Even if I load the next tape in the sequence mnaually, amcheck rejects it. How can I tell which tapes it will write too? dumpcycle 7 days runspercycle 5 tapecycle 15 tapes runtapes 2 So far, backups have only needed one tape per run so it seems to me that tapes from two weeks ago should no longer be active but amanda thinks they are. You need to decrease tapecycle. That says how many tapes you've got in rotation. You could lower it to 10 which would say that it can't rewrite a tape for 2 weeks. -- [EMAIL PROTECTED] - HMC UNIX Systems Manager My opinions are my own and probably don't represent anything anyway.
RE: Missing Tape has Backups Screwed
Oops. Do you have runtapes 2 ? I think you would have to mark all of the tapes that are in this run as no-reuse to get it to cycle around. Do you have runtapes * runspercycle = tapecycle? (ie, runtapes=2, runspercycle=5, tapecycle=10 or something like that?) I should have been more explicit about there needing to be spare tapes in the rotation for no-reuse to do what you would expect. From what I've seen on the list, it's a good idea to have tapecycle runtapes*runspercycle, for just this reason. You can temporarily pull a tape from the rotation without confusing AMANDA. (For example, I have runtapes=1, runspercycle=5, tapecycle=30, so I have six complete backup sets, one for each of the last six weeks. I picked those numbers based on the number of tape magazines I had for my changer.) Good Luck. -Original Message- From: Stephen Carville [mailto:[EMAIL PROTECTED]] Sent: Friday, February 08, 2002 4:21 PM To: Bort, Paul Cc: Amanda Users Subject: RE: Missing Tape has Backups Screwed On Fri, 8 Feb 2002, Bort, Paul wrote: - amadmin YOURCONFIG no-reuse MISSINGTAPE - - Then use the next tape in sequence. - - Just remember to mark it 'reuse' before it is next scheduled. Didn't work. $ amadmin daily1 tape The next Amanda run should go onto tape C83173 or a new tape. The next Amanda run should go onto tape C83174 or a new tape. $ amadmin daily1 no-reuse C83173 amadmin: marking tape C83173 as not reusable. $ amtape daily1 label C83174 amtape: scanning for tape with label C83174 amtape: slot 9: date 20020121 label C83174 (exact label match) amtape: label C83174 is now loaded. amdump reports *** A TAPE ERROR OCCURRED: [new tape not found in rack]. When I try 'amcheck daily1' it reports that _all_ the tapes are active. Even tapes that 'amtape daily1 info' says are not part of the current backup. -- Stephen Carville UNIX and Network Administrator DPSI (formerly Ace USA Flood Services) 310-342-3602 [EMAIL PROTECTED]
RE: Missing Tape has Backups Screwed
On Fri, 8 Feb 2002, Bort, Paul wrote: - Oops. Do you have runtapes 2 ? I think you would have to mark all of the - tapes that are in this run as no-reuse to get it to cycle around. - - Do you have runtapes * runspercycle = tapecycle? (ie, runtapes=2, - runspercycle=5, tapecycle=10 or something like that?) I should have been - more explicit about there needing to be spare tapes in the rotation for - no-reuse to do what you would expect. - - From what I've seen on the list, it's a good idea to have tapecycle - runtapes*runspercycle, for just this reason. You can temporarily pull a tape - from the rotation without confusing AMANDA. (For example, I have runtapes=1, - runspercycle=5, tapecycle=30, so I have six complete backup sets, one for - each of the last six weeks. I picked those numbers based on the number of - tape magazines I had for my changer.) - - Good Luck. Decreasing tapecycle from 15 to 10 did the trick. Thanks -- -- Stephen Carville UNIX and Network Administrator DPSI (formerly Ace USA Flood Services) 310-342-3602 [EMAIL PROTECTED]
Amanda tape index script
Before I spend a little time on this I thought I'd ask, a quick look through the archives did not seem to yield any results. I want a little shell/perl/python, etc. script that I can use to recreate an Amanda index. I have a couple of old Amanda installations that never used indexing. So now when I have to restore from them I end up mt'ing and dd'ing the headers one at a time to find the right one. What I would like is a little script that creates an index file for a tape with the header and fsf number. Anyone already have one or should I just hack something together? Thanks -- Sacha Panasuik Senior Systems Administrator TCENet Inc. - A Technology Company Reality is just a crutch for people who can't handle science fiction
closing tape: Input/output error
I've not seen the following error in 2.5 years of minding Amanda backups, nor have I found it in the archives: *** THE DUMPS DID NOT FINISH PROPERLY! *** A TAPE ERROR OCCURRED: [[closing tape: Input/output error]]. *** PERFORMED ALL DUMPS TO HOLDING DISK. THESE DUMPS WERE TO DISK. Flush them onto tape DLT07 or a new tape. Tonight's dumps should go onto 1 tape: DLT08. FAILURE AND STRANGE DUMP SUMMARY: shell da0s1f lev 0 FAILED [data write: Broken pipe] shell da0s1f lev 0 FAILED [out of tape] taper: FATAL syncpipe_get: w: unexpected EOF shell da0s1f lev 0 FAILED [dump to tape failed] I found the following in the server's messages file: Feb 8 07:25:48 storage /kernel: (sa0:bt0:0:1:0): CCB 0xcde9e300 - timed out Feb 8 07:25:50 storage /kernel: (sa0:bt0:0:1:0): CCB 0xcde9e300 - timed out Feb 8 07:25:50 storage /kernel: bt0: No longer in timeout Feb 8 07:25:50 storage /kernel: (sa0:bt0:0:1:0): failed to write terminating filemark(s) Feb 8 07:25:55 storage /kernel: (sa0:bt0:0:1:0): tape is now frozen- use an OFFLINE, REWIND or MTEOM command to clear this state. Now, this sort of looks like a bad tape. Or is the [out of tape] error true? If so, I thought amanda was supposed to be able to figure things so everything fits on a tape. I know I've seen amanda do incs when a full would have been too much for a tape. Regards, --Ruth Anne
Re: Amanda tape index script
I want a little shell/perl/python, etc. script that I can use to recreate an Amanda index. ... That would be most welcome (although please use sh (not bash) or Perl to keep in the spirit of what the rest of Amanda releases). However (you knew that was coming, right? :-) ... This is a much harder thing to do than you might think. If the image is created with GNU tar, and you have GNU tar on the tape machine, then it's not too bad. Otherwise, you have to send the image back to the original machine (or a machine with the same restore software) to create the index, then grab that output back on the server and run it through the appropriate sed to create the file. There's a reason nobody has fought this battle yet :-). Sacha Panasuik John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: Can you change /tmp/amanda to a more permanent directory?
... does Amanda recreate its /tmp/amanda directory, or expect it to already be there? ... It recreates it in lots of places (pretty much any time a program starts), but if Amanda is running and stores a file there that the system removes before the program is done with it, that's the problem. Personally, I put my tmp files elsewhere for this very reason. I keep 10 days worth of debug files and on the worst configuration I have it's only 7 MBytes. Deb Baddorf John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: closing tape: Input/output error
*** A TAPE ERROR OCCURRED: [[closing tape: Input/output error]]. This comes from a couple of places in Amanda, but the most likely is that Amanda was all done and was just finishing up. The tape is *probably* OK, but I wouldn't bet everything on it. taper: FATAL syncpipe_get: w: unexpected EOF I think this is fixed in the current sources. ... is the [out of tape] error true? ... That error means Amanda got an error writing to the last tape you told it it could use (runtapes) so it (Amanda) is out of tape. It does not literally mean the error on this particular tape was that it ran out of space. I found the following in the server's messages file: Feb 8 07:25:48 storage /kernel: (sa0:bt0:0:1:0): CCB 0xcde9e300 - timed out Now, this sort of looks like a bad tape. ... Or some other kind of hardware problem. Timeouts are a bad thing. ... I thought amanda was supposed to be able to figure things so everything fits on a tape. ... It takes all the estimated dump sizes and fits them into the total amount of space you told it it could have on tape (runtapes * the tapetype size). If the real dump is larger than the estimate, or the tape is shorter (including getting an error), then there isn't a lot Amanda can do about that. If you're using compression, that can also shorten the available tape space if the dump does not compress as well as expected. You left out one important line from the E-mail. The one in the NOTES section that shows how much data taper wrote before it got into trouble. If that's close to your tape size, then you may have just banged into the end. If it's way short, then it's a hardware or media problem. --Ruth Anne John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
RE: Amanda tape index script
What would the advantages/disadvantages be of using GNU tar for all your backup needs? Is it less efficient than vendor dump utilities? Unwanted side effects? B -Original Message- From: John R. Jackson [mailto:[EMAIL PROTECTED]] Sent: Friday, February 08, 2002 2:51 PM To: Sacha Panasuik Cc: [EMAIL PROTECTED] Subject: Re: Amanda tape index script I want a little shell/perl/python, etc. script that I can use to recreate an Amanda index. ... That would be most welcome (although please use sh (not bash) or Perl to keep in the spirit of what the rest of Amanda releases). However (you knew that was coming, right? :-) ... This is a much harder thing to do than you might think. If the image is created with GNU tar, and you have GNU tar on the tape machine, then it's not too bad. Otherwise, you have to send the image back to the original machine (or a machine with the same restore software) to create the index, then grab that output back on the server and run it through the appropriate sed to create the file. There's a reason nobody has fought this battle yet :-). Sacha Panasuik John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: Q. R.e. smb amanda
In my disklist: adv2.kellogg.nwu.edu//adv.kellogg.nwu.edu/web/cos nocomp-user-gnu tar In my /etc/amandapass: //adv.kellogg.nwu.edu/web/cos amanda%xx I think that should be: //adv.kellogg.nwu.edu/web amanda%xx The first column is the *share* name, not the disklist name. Amanda Backup Client Hosts Check Client check: 1 host checked in 1.211 seconds, 0 problems found (brought to you by Amanda 2.4.3b2) But here's the other thing - it says 1 host checked, but I've got TWO hosts under my disklist ... so whats up with that... Whatever. No, you only have one client, adv2.kellogg.nwu.edu. The PC (adv) is not a host as far as Amanda is concerned. It's a disk that Amanda thinks is connected to adv2. So, I try running a dump anyway, and i get this: ... FAILURE AND STRANGE DUMP SUMMARY: adv2.kello //adv.kellogg.nwu.edu/web lev 0 FAILED [can't switch to incremental dump] So, don't keep us in suspense, what's in the FAILED AND STRANGE DUMP DETAILS portion of the E-mail??? :-) What's in the corresponding /tmp/amanda/sendbackup*debug file on adv2? John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: Amanda tape index script
Hmm. Its funny how an idea is completely lucid in my brain and yet gets mixed up between brain and keyboard. :) What I mean by index is something like this: Position: host: file system not a complete index of each file available to be restored. So that what I'm currently calling an index would be a file with as many lines as there are dumps on a tape. There is that more confusing :) John R. Jackson wrote: I want a little shell/perl/python, etc. script that I can use to recreate an Amanda index. ... That would be most welcome (although please use sh (not bash) or Perl to keep in the spirit of what the rest of Amanda releases). -- Sacha Panasuik, MCSE Senior Systems Administrator TCENet Inc. - A Technology Company Reality is just a crutch for people who can't handle science fiction
Re: Amanda tape index script
What would the advantages/disadvantages be of using GNU tar for all your backup needs? Is it less efficient than vendor dump utilities? Unwanted side effects? In general, it's significantly slower, and touches the read dates on all your files. On the other hand, it's possible to break up a large filesystem into chunks that are easier for Amanda to swallow.
Re: Amanda tape index script
Hmm. Its funny how an idea is completely lucid in my brain and yet gets mixed up between brain and keyboard. :) Tell me about it :-)! What I mean by index is something like this: Position: host: file system not a complete index of each file available to be restored. ... Oh, I see. Will amtoc do what you want? It only uses the log.MMDD.NN files. Here's a sample: # Server:/partition date level size[Kb] 0 B00100/acmaint: 20011230 - - 1 pythia.cc.purdue.edu:/export/home/b 20011230 1 63 2 catbert.cc.purdue.edu:/export/home/a 20011230 0 31 3 ledger.cue.purdue.edu:/var/adm/acmlog 20011230 1 3615 ... Or, if they are old tapes that you don't have the log.MMDD.NN files for any more and so can't run amtoc against: mt rewind amrestore $TAPE no-such-host This scans the tape looking for a client named no-such-host. As it wanders along it reports everything else it sees. Not as nicely parseable as your idea, but that's what sed is for :-). Sacha Panasuik John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: closing tape: Input/output error
On Fri, 8 Feb 2002, John R. Jackson wrote: *** A TAPE ERROR OCCURRED: [[closing tape: Input/output error]]. This comes from a couple of places in Amanda, but the most likely is that Amanda was all done and was just finishing up. The tape is *probably* OK, but I wouldn't bet everything on it. Depending on the the comments after I fill in some blanks, it'll probably get repleaced anyway. taper: FATAL syncpipe_get: w: unexpected EOF I think this is fixed in the current sources. Noted. Feb 8 07:25:48 storage /kernel: (sa0:bt0:0:1:0): CCB 0xcde9e300 - timed out Now, this sort of looks like a bad tape. ... Or some other kind of hardware problem. Timeouts are a bad thing. Other staff put the tape drive on a buslogic scsi card last night, to hopefully solve a crash problem we were having. Could it be causing the problem? (I probably should have mentioned that in the initial email.) If our problem happens again tonight, we'll probably know where the issue lies. You left out one important line from the E-mail. The one in the NOTES section that shows how much data taper wrote before it got into trouble. If that's close to your tape size, then you may have just banged into the end. If it's way short, then it's a hardware or media problem. Beg pardon, here's the missing bit: taper: tape NOMDLT06 kb 3012224 fm 17 writing file: Input/output error The tapes are AIT-1 25/50M, no compression that I can tell. Thanks, --Ruth Anne
Re: follow up
What I'm looking to do is something somwhat automatic - so I don't have to do this every four days. I.e. is there a way to tell amdump to overwrite a backup tape? Ummm, yes. :-). Your tapecycle value tells Amanda how many tapes it has to work with. Don't lie to Amanda :-). If you have 10 tapes (for instance), set tapecycle to 10. After the 10th one is used, it will overwrite the first. If so, then I wouldn't have to do all this amlabel crap. ... You should only have to amlabel your tapes once. If you're having to do it more than once (other than initial screwing around and testing) something else is wrong. Also, is there a way to get amanda to make .tar.gz's instead of its own format? ... No. But you can easily get back .tar.gz's from the tapes with amrestore (look at the -c option). In addition, is there a way tt amanda to run several amandas at once? I'm backing up stuff thats on a slow connection, so I've got plenty of bandwith.. Sine amanda does stuff sequentially, I'm stuck waiting 20 minutes for some file to backup and im only using like uhm, 20k/s... So I could be running 10 other amanda backups at teh same time Not sure what you mean by sequentially. If you set inparallel to 8 (just to pick a number), Amanda (amdump) will run up to eight backups at the same time, subject to other constraints (maxdumps controls the number of dumps a single client can do, for instance). Do I just make multiple confs and put them all in my crontabs? You could, but I'm not sure you need to resort to this yet. cosimo John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: closing tape: Input/output error
Other staff put the tape drive on a buslogic scsi card last night, to hopefully solve a crash problem we were having. Could it be causing the problem? ... Aha! The truth finally comes out :-). Sure, it could be causing the problem. Any number of things could have gone wrong during that swap -- too much or not enough termination, cable not seated well enough, bent pins, bad controller, full reset needed, etc. taper: tape NOMDLT06 kb 3012224 fm 17 writing file: Input/output error The tapes are AIT-1 25/50M, no compression that I can tell. So you weren't anywhere close to the tape capacity, which puts this squarely in the hardware error (as compared to end of media) problem arena. --Ruth Anne John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: amdump question
I have problem with amdump: host which I try to dump does not respond (by time-out) but it is alive. ... FAIL planner local /dev/rsd0a 0 [Request to local timed out.] Did you run amcheck? What did it say? If you get timeouts there, look at the FAQ at www.amanda.org. It has two articles that go over this problem in depth. Mary John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: amcheck
When I run su amanda -c amcheck daily the backup server times out with the selfcheck request timed out error. Believe it or not, this just started happening, and I _haven't_ changed anything. Yeah, yeah. We've heard all that before :-). The funny part is this: amdump works fine and runs nightly! It backs-up the server, even tho amcheck says it's down. Amcheck and amdump (dumper) use **exactly** the same access mechanisms to the clients. So you're clearly just imagining this. It can't be happening :-). When you run amcheck, does a new /tmp/amanda/amandad*debug file get created (or the existing one updated)? What's in it? In particular, what are the stop and start times (first and last lines)? Ditto for selfcheck*debug. ... I doubled the etimeout from 300 to 600, but amcheck still times out. That makes sense. Etimeout is for estimates done during amdump. You want to crank up ctimeout (for check timeout). Brad John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: backups still failing.
where doing Daily backups to DDS3 tapes with a block size of 4096, which seem to work sometimes. more often than if I use a block size of 0. I'm amazed that works unless you re-amlabel the tapes whenever you change the blocksize. Going either direction without doing so should cause errors. and where doing Monthly backups to DDS4 tapes with a block size of 4096, and they completely fail with the error messages below. Are you certain those tapes were created with 4K? Set the block size back to zero (for the moment) and do this: mt rewind dd if=$TAPE bs=32k of=/tmp/first-record count=1 mt rewind dd if=$TAPE bs=32k of=/tmp/first-file See how big /tmp/first-record is. That's the block size the tape was written at. If it's 4K, the second dd should tell you it read 0+8 records (no full 32K blocks, 8 partials), and /tmp/first-file should be 32K. John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: closing tape: Input/output error
On Fri, 8 Feb 2002, John R. Jackson wrote: Other staff put the tape drive on a buslogic scsi card last night, to hopefully solve a crash problem we were having. Could it be causing the problem? ... Aha! The truth finally comes out :-). Hey, I didn't do this so I didn't remember until after I sent the original email. :) Sure, it could be causing the problem. Any number of things could have gone wrong during that swap -- too much or not enough termination, cable not seated well enough, bent pins, bad controller, full reset needed, etc. taper: tape NOMDLT06 kb 3012224 fm 17 writing file: Input/output error The tapes are AIT-1 25/50M, no compression that I can tell. So you weren't anywhere close to the tape capacity, which puts this squarely in the hardware error (as compared to end of media) problem arena. OK, I thought as much by this time, with your help. :) I'll be watching things tonight... if there's another failure, I can pretty much pin it on the new scsi controller. Thanks, --Ruth Anne
Re: ERROR: new tape not found in rack
I started to use amanda yesterday. I am beginer about amanda. Welcome! When su amanda -c amcheck Daily is running, I am getting the following ERROR. ERROR: new tape not found in rack. (expecting a new tape) What version of Amanda are you using? Did you run amlabel on all your tapes? Which tape changer did you put in amanda.conf? Weren't there other messages from amcheck before the ERROR? What were they? What does amtape Daily show have to say? Masafumi Hikawa John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: closing tape: Input/output error
... if there's another failure, I can pretty much pin it on the new scsi controller. Or the termination. Or the cable. Etc. BTW, your return address ([EMAIL PROTECTED]) does not work. It bounces with User unknown. --Ruth Anne John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: [Amanda-users] Re: backups still failing.
On Fri, Feb 08, 2002 at 07:31:08PM -0500, John R. Jackson wrote: where doing Daily backups to DDS3 tapes with a block size of 4096, which seem to work sometimes. more often than if I use a block size of 0. I'm amazed that works unless you re-amlabel the tapes whenever you change the blocksize. Going either direction without doing so should cause errors. and where doing Monthly backups to DDS4 tapes with a block size of 4096, and they completely fail with the error messages below. Are you certain those tapes were created with 4K? Set the block size back to zero (for the moment) and do this: mt rewind dd if=$TAPE bs=32k of=/tmp/first-record count=1 mt rewind dd if=$TAPE bs=32k of=/tmp/first-file See how big /tmp/first-record is. That's the block size the tape was written at. If it's 4K, the second dd should tell you it read 0+8 records (no full 32K blocks, 8 partials), and /tmp/first-file should be 32K. its not happy: root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn status SCSI 2 tape drive: File number=0, block number=0, partition=0. Tape block size 4096 bytes. Density code 0x26 (DDS-4 or QIC-4GB). Soft error count since last status=0 General status bits on (4101): BOT ONLINE IM_REP_EN root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn setblk 0 root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn status SCSI 2 tape drive: File number=0, block number=0, partition=0. Tape block size 0 bytes. Density code 0x26 (DDS-4 or QIC-4GB). Soft error count since last status=0 General status bits on (4101): BOT ONLINE IM_REP_EN root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn rewind root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn of=/tmp/first-record count=1 dd: reading `/dev/scsi/host0/bus0/target15/lun0/mtn': Cannot allocate memory 0+0 records in 0+0 records out root@bast:~# rm /tmp/first-record root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn of=/tmp/first-filedd: reading `/dev/scsi/host0/bus0/target15/lun0/mtn': Cannot allocate memory 0+0 records in 0+0 records out root@bast:~# ls -la /tmp/first-record -rw-r--r--1 root root0 Feb 9 12:46 /tmp/first-record root@bast:~# ls -la /tmp/first-file -rw-r--r--1 root root0 Feb 9 12:46 /tmp/first-file root@bast:~# but with it back at 4096: root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn setblk 4096 root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn status SCSI 2 tape drive: File number=0, block number=2, partition=0. Tape block size 4096 bytes. Density code 0x26 (DDS-4 or QIC-4GB). Soft error count since last status=0 General status bits on (101): ONLINE IM_REP_EN root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn rewind root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn of=/tmp/first-record count=1 1+0 records in 1+0 records out root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn of=/tmp/first-file 0+0 records in 0+0 records out root@bast:~# ls -la /tmp/first-record -rw-r--r--1 root root 512 Feb 9 12:48 /tmp/first-record root@bast:~# ls -la /tmp/first-file -rw-r--r--1 root root0 Feb 9 12:48 /tmp/first-file root@bast:~#
Re: [Amanda-users] Re: backups still failing.
its not happy: ... root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn of=/tmp/first-record count=1 dd: reading `/dev/scsi/host0/bus0/target15/lun0/mtn': Cannot allocate memory You forgot the bs=32k. What happens if you add that? And you need to do a rewind between the two dd's. John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: [Amanda-users] Re: backups still failing.
On Fri, Feb 08, 2002 at 10:15:32PM -0500, John R. Jackson wrote: You forgot the bs=32k. What happens if you add that? And you need to do a rewind between the two dd's. sorry, this is better: root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn status SCSI 2 tape drive: File number=0, block number=0, partition=0. Tape block size 4096 bytes. Density code 0x26 (DDS-4 or QIC-4GB). Soft error count since last status=0 General status bits on (4101): BOT ONLINE IM_REP_EN root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn setblk 0 root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn status SCSI 2 tape drive: File number=1, block number=0, partition=0. Tape block size 0 bytes. Density code 0x26 (DDS-4 or QIC-4GB). Soft error count since last status=0 General status bits on (8101): EOF ONLINE IM_REP_EN root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn rewind root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn bs=32k of=/tmp/first-record count=1 0+1 records in 0+1 records out root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn rewind root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn bs=32k of=/tmp/first-file 0+7 records in 0+7 records out root@bast:~# ls -la /tmp/first-* -rw-r--r--1 root root32768 Feb 9 14:44 /tmp/first-file -rw-r--r--1 root root32768 Feb 9 14:43 /tmp/first-record root@bast:~#
Re: [Amanda-users] Re: backups still failing.
root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn bs=32k of=/tmp/first-record count=1 0+1 records in 0+1 records out root@bast:~# mt -f /dev/scsi/host0/bus0/target15/lun0/mtn rewind root@bast:~# dd if=/dev/scsi/host0/bus0/target15/lun0/mtn bs=32k of=/tmp/first-file 0+7 records in 0+7 records out root@bast:~# ls -la /tmp/first-* -rw-r--r--1 root root32768 Feb 9 14:44 /tmp/first-file -rw-r--r--1 root root32768 Feb 9 14:43 /tmp/first-record This is odd. The output from dd about record counts does not match the final size of the files. Could you try this again and make sure you remove the /tmp files first. Also, try it with bs=64k. John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]