[CRIU] [Users] OVZ 7 beta - ploop backup fully functional?

[Kir Kolyshkin]

On 11/04/2015 10:16 AM, jjs - mainphrame wrote:
> Greetings,
>
> I'm still running OVZ 7 here on 2 machines, one running beta and the 
> other running factory. vzctl snapshot hasn't worked on either one, but 
> I test at intervals to check the status. The factory machine got some 
> updates today, and so I tried again, and things seemed to get a bit 
> farther before it failed. An excerpt from near the end of the dump log 
> contained some clues:
>
> (00.159708) Dumping ghost file for fd 26 id 0x7c
> (00.159712) Error (files-reg.c:422): Can't dump ghost file 
> /usr/lib64/libnss3.so;56314442 of 1220096 size, increase limit

For this particular thing there might be a couple of issues filed and 
being worked on, please
search Jira.

The ghost file limit is set in CRIU by

   --ghost-limit size    specify maximum size of deleted file contents 
to be carried inside an image file

Although I'm not aware how to pass this flag from vzctl.

Also I saw some on-going work about migrating such deleted files outside 
of image (to improve migrate frozen time).
Again, it's either on Jira or on devel@ list (or perhaps both).

> (00.159718) Error (cr-dump.c:1314): Dump mappings (pid: 14025) failed 
> with -1
>
> Are there any environment variables or command line switches I could 
> apply which could affect this "limit" hinted at above?
>
> BTW - criu doesn't seem entirely happy with the ovz 7 kernel:
>
> [root at annie ~]# uname -a
> Linux annie.mainphrame.net <http://annie.mainphrame.net> 
> 3.10.0-229.7.2.vz7.9.1 #1 SMP Wed Oct 21 17:55:13 MSK 2015 x86_64 
> x86_64 x86_64 GNU/Linux
> [root at annie ~]# criu check
> Error (cr-check.c:602): Kernel doesn't support PTRACE_O_SUSPEND_SECCOMP

This shouldn't bother you as far as I understand as seccomp is not being 
used.

> Error (cr-check.c:719): fdinfo doesn't contain the lock field
> Error (cr-check.c:749): CLONE_PARENT | CLONE_NEWPID don't work together

For this I hope CRIU guys can shed more light (ccing CRIU list)

> [root at annie ~]#
>
> Am I tilting at windmills here, or is there some expectation that any 
> of this should be working?
>
> Regards,
>
> Joe
>
>
>
>
>
>
>
> On Wed, Oct 28, 2015 at 2:31 PM, jjs - mainphrame <jjs at mainphrame.com 
> <mailto:jjs at mainphrame.com>> wrote:
>
>     Greetings,
>
>     I've been running some test containers on OVZ7 beta, and, while I
>     realize not all functionality is in place yet, they have been so
>     dependable that I'm starting to depend on them.
>
>     So, understandably, I was looking at backups, and my first tries
>     with vzctl snapshot failed. I don't want to spend a lot of time
>     troubleshooting something that's not yet known to be working, but
>     in case it is, has anyone else had better luck?
>
>
>     Here is the result of my attempt to make a backup:
>
>     [root at hachi vz]# vzctl snapshot 1001
>     Creating snapshot {57c8aa0d-1c04-4469-9d1e-c95fab04bd01}
>     Storing /local/vz/private/1001/Snapshots.xml.tmp
>     Setting up checkpoint...
>     Failed to checkpoint the Container
>     All dump files and logs were saved to
>     /local/vz/private/1001/dump/{57c8aa0d-1c04-4469-9d1e-c95fab04bd01}.fail
>     Checkpointing failed
>     Failed to create snapshot
>     [root at hachi vz]#
>
>
>     The last few lines of the dump log contain these clues:
>
>     (00.570576) Error (proc_parse.c:404): Can't handle non-regular
>     mapping on 8366's map 7f2164d63000
>     (00.570582) Error (cr-dump.c:1487): Collect mappings (pid: 8366)
>     failed with -1
>     (00.570746) Unlock network
>     (00.570751) Running network-unlock scripts
>
>
>     There is no pid 8366 in the CT but on the host that pid
>     corresponds to the mysql instance running in the container that
>     I'm trying to snapshot:
>
>     [root at hachi vz]# ps -ef|grep 8366
>     27          8366    7098  0 14:03 ?        00:00:00
>     /usr/libexec/mysqld --basedir=/usr --datadir=/var/lib/mysql
>     --plugin-dir=/usr/lib64/mysql/plugin
>     --log-error=/var/log/mariadb/mariadb.log
>     --pid-file=/var/run/mariadb/mariadb.pid
>     --socket=/var/lib/mysql/mysql.sock
>
>     So, is this something that ought to work?
>
>     Regards,
>
>     Joseph
>
>
>
>
> _______________________________________________
> Users mailing list
> Users at openvz.org
> https://lists.openvz.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvz.org/pipermail/criu/attachments/20151104/5f1df77f/attachment.html>