Re: Amanda Runs Hanging
Jim Summers wrote:
Hello All,
I started having my nightly backups hang. The backups have been working
fine until Thursday of last week.
The server is amanda-2.5.2p1 on rhel4 and the client that I believe is
involved is amanda-2.6.0p2 on centos5. Their are other clients and they
are amanda-2.5.2p1 on rhel4 or fc8 also.
It seems to not be completing the estimates or ever coming out of the
planner phase. It never actually dumps or does any backup. The next
day when the scheduled amcheck runs I get an email saying that the
amdump or amflush is running and I need to run amcleanup if needed.
Running amcleanup returns "Results Missing" for all of the dle's for
each of the hosts.
So I am not sure where the issue is. I have increased the etimeout from
300 to 3600 and the results seem to be the same. I also re-compiled the
client and server to only use ipv4. Still same results.
I found the following error in the amandad.x.debug file on the client:
>
1231933529.123236: amandad: dgram_send_addr(addr=0x5b524d0,
dgram=0x2af76db07d08)
1231933529.123257: amandad: (sockaddr_in *)0x5b524d0 = { 2, 889,
129.15.11.211 }
1231933529.123274: amandad: dgram_send_addr: 0x2af76db07d08->socket = 0
1231933835.099966: amandad: /usr/local/libexec/amanda/sendsize timed out
waiting for REP data
1231933835.100027: amandad: sending NAK pkt:
<
ERROR timeout on reply pipe
>
1231933835.100056: amandad: dgram_send_addr(addr=0x5b524d0,
dgram=0x2af76db07d08)
1231933835.100076: amandad: (sockaddr_in *)0x5b524d0 = { 2, 889,
129.15.11.211 }
1231933835.100093: amandad: dgram_send_addr: 0x2af76db07d08->socket = 0
1231933835.100219: amandad: security_close(handle=0x5b52490,
driver=0x2af76dafe3e0 (BSD))
1231933864.098431: amandad: pid 19716 finish time Wed Jan 14 05:51:04 2009
==
and then at the end of the planner.xx.debug file on the tape server
I see:
planner: time 11076.021: (sockaddr_in *)0x627530 = { 2, 10080,
129.15.11.173 }
planner: time 11092.842: dgram_recv(dgram=0x617544, timeout=0,
fromaddr=0x627530)
planner: time 11092.842: (sockaddr_in *)0x627530 = { 2, 10080,
129.15.11.173 }
planner: time 21324.684: dgram_recv(dgram=0x617544, timeout=0,
fromaddr=0x627530)
planner: time 21324.705: (sockaddr_in *)0x627530 = { 2, 10080,
129.15.11.173 }
planner: time 21630.663: dgram_recv(dgram=0x617544, timeout=0,
fromaddr=0x627530)
planner: time 21630.663: (sockaddr_in *)0x627530 = { 2, 10080,
129.15.11.173 }
==
The odd thing there is there is not a dgram_recv for the last entry into
the log. From there everything just seems to stop.
Any ideas or suggestions?
Please let me know and I can provide more debug if needed.
TIA
I found a thread that mentioned that possibly the estimate was timing out and
to try using the calcsize program for the estimate. I made that switch and
the amdump run is now running.
Thanks
--
Jim Summers
School of Computer Science-University of Oklahoma
-
Re: RPM files for openSUSE 11.1
Hi all, Julian's email reminded me to check the Build Service archive, and there I found amanda-2.6.0.2-2.1.src.rpm, apparently for o-suse 11.1. I decided to do a rmpbuild on that file, and it worked until the end and created 3 rpm's: - amanda-2.6.0.2-2.1.i586.rpm - amanda-backup_client-2.6.0.2-2.1.i586.rpm - amanda-backup_server-2.6.0.2-2.1.i586.rpm in the /usr/src/packages/RPMs/i586 directory. Actually, there were also binary rpms for this amanda version, but these are very much different in size compared with my generated files. Installation of my backup-server rpm was also ok, except that it did all of this under a created user "amanda" and not "amandabackup". Apart from changing all files manually, is there a way to create binary rpms with generate "amandabackup" user? But I cannot make sense of Julian's last para's: that is beyond my capabilities. Regards, Charles On Thu, 15 Jan 2009 02:44:25 + "Julian C. Dunn" wrote: > I have an amanda-2.6.0 package on OpenSUSE Build Service that I've > been messing around with. > > https://build.opensuse.org/project/show?project=home:dunnj > > I cribbed the spec file from lmich on OBS and fixed some of the % > Buildrequires. But OBS should be building packages for both 11.0 and > 11.1. > > Feel free to hack on it yourself. > > - Julian >
Re: RPM files for openSUSE 11.1
On Wed, Jan 14, 2009 at 7:07 AM, Charles Stroom
wrote:
> fiume: # rpmbuild --rebuild amanda-2.6.0p2-1.noarch.src.rpm
> Installing amanda-2.6.0p2-1.noarch.src.rpm
> warning: user buildslave does not exist - using root
> warning: group buildslave does not exist - using root
> warning: user buildslave does not exist - using root
> warning: group buildslave does not exist - using root
> PKG_CONFIG_PATH=/opt/gnome/lib/pkgconfig:/opt/gnome/share/pkgconfigerror:
> parse error in expression
> error: /usr/src/packages/SPECS/amanda.spec:113: parseExpressionBoolean
> returns -1
> error: Name field must be present in package: (main package)
> error: Version field must be present in package: (main package)
> error: Release field must be present in package: (main package)
> error: Summary field must be present in package: (main package)
> error: Group field must be present in package: (main package)
> error: License field must be present in package: (main package)
If you don't mind continuing to work on this, I can point you toward
the problem line. The part that begins with
34 # Define which Distribution we are building:
35 # Try to detect the distribution we are building:
tries to determine which of the recognized redhat variants we're
building for. Since we don't do OpenSUSE, we don't recognize it
specifically. As Ingo pointed out, the macros to use would be in your
suse_macros file and friends. The line that's giving the error is:
112 # Detect Suse variants. Suse gives us some nice macros in their rpms
113 %if %{_vendor} == "suse"
114 %if %{suse_version} == 910
so I'm guessing that _vendor isn't defined.
Basically, all I need is for someone or a group (hi, Charles and
Ingo!) reasonably familiar with the macros available on OpenSUSE to
add an extra stanza to this long set of conditionals that can
recognize OpenSUSE releases. We'll happily merge that extra stanza
into the distribution, and voila! you can rpmbuild --rebuild RPMs as
often as you like :)
Regarding "amanda" vs "amandabackup" and other variants between
"zmanda" RPMs and distro RPMs -- as the authors of the software, we
try to standardize things *across* Linux distros, since in many cases
Amanda users back up a wide range of machine flavors, and it's a lot
easier to do so when they all have similar Amanda installs[1]. So I'd
love to see the packages shipped with distributions and from sites
like rpmforge start to adopt the standards represented in our SPECs,
but I realize that this often creates upgrade headaches for
distrubution maintainers. Such is life..
Dustin
[1] Oh, and it makes it a lot easier to support too!
--
Storage Software Engineer
http://www.zmanda.com
