Full backup - VM failed in first attempt but successful in second attempt

rtjdamen

@hbogner sorry i am sleeping, monday morning ;-). mixed it up with a different thread.

hbogner

@olivierlambert said in Full backup - VM failed in first attempt but successful in second attempt:

Also check your Dom0 memory, I have the feeling the zstd compression is failing and leaving the export opened until XO timeouts

When I disable compression backup is successful in the first run, but it lasts longer. This VM is finished, but larger ones are still at it, they'll finish tomorrow, estimated by XO in tasks menu.

hbogner

With zstd compression backup fails on the first run, lasts 11 hours, Transfer: 1.81 TiB.

With no compression backup runs ok in the first run, lasts 15 hours, Transfer: 5.12 TiB.

I'd like to keep it compressed in the long run

Upgraded to latest: "Xen Orchestra(commit 74e6f), Master(commit 74e6f), Your Xen Orchestra is up to date"

Started new compressed backup and it failed again:

root@backup:~# grep "xo-server" /var/log/syslog
2024-09-13T09:24:17.518727+02:00 backup xo-server[243615]: 2024-09-13T07:24:17.516Z xo:backups:worker INFO starting backup
2024-09-13T11:11:56.199073+02:00 backup xo-server[243615]: 2024-09-13T09:11:56.195Z xo:backups:worker WARN possibly unhandled rejection {
2024-09-13T11:11:56.199261+02:00 backup xo-server[243615]:   error: BodyTimeoutError: Body Timeout Error
2024-09-13T11:11:56.199310+02:00 backup xo-server[243615]:       at Timeout.onParserTimeout [as callback] (/opt/xen-orchestra/node_modules/undici/lib/dispatcher/client-h1.js:626:28)
2024-09-13T11:11:56.199337+02:00 backup xo-server[243615]:       at Timeout.onTimeout [as _onTimeout] (/opt/xen-orchestra/node_modules/undici/lib/util/timers.js:22:13)
2024-09-13T11:11:56.199362+02:00 backup xo-server[243615]:       at listOnTimeout (node:internal/timers:581:17)
2024-09-13T11:11:56.199387+02:00 backup xo-server[243615]:       at process.processTimers (node:internal/timers:519:7) {
2024-09-13T11:11:56.199408+02:00 backup xo-server[243615]:     code: 'UND_ERR_BODY_TIMEOUT'
2024-09-13T11:11:56.199428+02:00 backup xo-server[243615]:   }
2024-09-13T11:11:56.199450+02:00 backup xo-server[243615]: }
2024-09-13T11:11:57.748715+02:00 backup xo-server[243615]: 2024-09-13T09:11:57.745Z xo:backups:AbstractVmRunner WARN writer step failed {
2024-09-13T11:11:57.748812+02:00 backup xo-server[243615]:   error: BodyTimeoutError: Body Timeout Error
2024-09-13T11:11:57.748837+02:00 backup xo-server[243615]:       at Timeout.onParserTimeout [as callback] (/opt/xen-orchestra/node_modules/undici/lib/dispatcher/client-h1.js:626:28)
2024-09-13T11:11:57.748858+02:00 backup xo-server[243615]:       at Timeout.onTimeout [as _onTimeout] (/opt/xen-orchestra/node_modules/undici/lib/util/timers.js:22:13)
2024-09-13T11:11:57.748878+02:00 backup xo-server[243615]:       at listOnTimeout (node:internal/timers:581:17)
2024-09-13T11:11:57.748898+02:00 backup xo-server[243615]:       at process.processTimers (node:internal/timers:519:7) {
2024-09-13T11:11:57.748917+02:00 backup xo-server[243615]:     code: 'UND_ERR_BODY_TIMEOUT'
2024-09-13T11:11:57.748937+02:00 backup xo-server[243615]:   },
2024-09-13T11:11:57.748961+02:00 backup xo-server[243615]:   step: 'writer.run()',
2024-09-13T11:11:57.748981+02:00 backup xo-server[243615]:   writer: 'FullRemoteWriter'
2024-09-13T11:11:57.749000+02:00 backup xo-server[243615]: }

olivierlambert

I think the issue might not be XO but XCP-ng with zstd.

hbogner

@olivierlambert you may be right, while digging thru logs on xcpng I found following in xensource.log:

Sep 13 11:18:22 clientxcpng xapi: [ warn||4289975 HTTPS <--IP ADRESS OF XO SERVER->:::80|[XO] VM export R:cd1a9a8d151f|pervasiveext] finally: Error while running cleanup after failure of main function: (Failure "Compression via zcat failed: exit code 25")
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] [XO] VM export R:cd1a9a8d151f failed with exception Server_error(CLIENT_ERROR, [ INTERNAL_ERROR: [ Unix.Unix_error(Unix.EPIPE, "write", "") ] ])
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] Raised Server_error(CLIENT_ERROR, [ INTERNAL_ERROR: [ Unix.Unix_error(Unix.EPIPE, "write", "") ] ])
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 1/18 xapi Raised at file ocaml/xapi/stream_vdi.ml, line 129
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 2/18 xapi Called from file ocaml/xapi/stream_vdi.ml, line 263
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 3/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 4/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 5/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 6/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 7/18 xapi Called from file ocaml/xapi/stream_vdi.ml, line 240
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 8/18 xapi Called from file list.ml, line 110
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 9/18 xapi Called from file ocaml/xapi/export.ml, line 653
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 10/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 11/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 12/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 13/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 14/18 xapi Called from file ocaml/xapi/server_helpers.ml, line 92
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 15/18 xapi Called from file ocaml/xapi/server_helpers.ml, line 113
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 16/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 17/18 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 18/18 xapi Called from file lib/backtrace.ml, line 177
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace]
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] VM.export D:1c0a6106f079 failed with exception Server_error(CLIENT_ERROR, [ INTERNAL_ERROR: [ Unix.Unix_error(Unix.EPIPE, "write", "") ] ])
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] Raised Server_error(CLIENT_ERROR, [ INTERNAL_ERROR: [ Unix.Unix_error(Unix.EPIPE, "write", "") ] ])
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace] 1/1 xapi Raised at file (Thread 4289975 has no backtrace table. Was with_backtraces called?, line 0
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||backtrace]
Sep 13 11:18:22 clientxcpng xapi: [error||4289975 :::80||xapi_http] Unhandled Api_errors.Server_error(CLIENT_ERROR, [ INTERNAL_ERROR: [ Unix.Unix_error(Unix.EPIPE, "write", "") ] ])

Is this any help?

olivierlambert

@hbogner said in Full backup - VM failed in first attempt but successful in second attempt:

Compression via zcat failed: exit code 25

Yes, it is useful. I wonder if you have enough RAM or disk space in your Dom0

hbogner

@hbogner said in Full backup - VM failed in first attempt but successful in second attempt:

@olivierlambert Dom0 memory is set to 12GiB of 256GiB available on the host.

12GB should be enough for 9 VM's.

Disk should also be ok:

Filesystem          Size  Used Avail Use% Mounted on
devtmpfs            5.9G   32K  5.9G   1% /dev
tmpfs               5.9G  532K  5.9G   1% /dev/shm
tmpfs               5.9G   11M  5.9G   1% /run
tmpfs               5.9G     0  5.9G   0% /sys/fs/cgroup
/dev/md127p1         18G  2.1G   15G  13% /
xenstore            5.9G     0  5.9G   0% /var/lib/xenstored
/dev/md127p5        3.9G  337M  3.3G  10% /var/log
tmpfs               1.2G     0  1.2G   0% /run/user/0

This is a Windows VM, with separate 100GB disk just for pagefile, could that be an issue while compressing?

olivierlambert

I don't know but I'm suspecting something going bad with ZSTD compression. Can you monitor your disk space during the export?

hbogner

@olivierlambert Tested it just now, no disk usage change on host(sample interval was 1 minute).
8 VM's finish ok, just this 1 VM fails on first run.

hbogner

@olivierlambert for now I "fixed" the problem by isolating this VM into separate backup job which doesn't overlap with any other backups and it finishes without error.

We'll try it again(smart mode VM group select) in a few months to see if new updates change anything.

olivierlambert

Even with compression? So it could be the dom0 overloaded with too many zstd pipes or something, interesting feedback.

hbogner

@olivierlambert yes, with compression.

In other 8 VMs there are larger VMs(larger diskd, more disks, more ram, more cpu), so export has higher load than this one VM, but only this one fails if it runs with concurrency more than 1.

Compression is a requirement for this situation because backups are synced to remote location over internet.

olivierlambert

I understand your requirements, I'm trying to pinpoint the issue. For now, everything converge on dom0 load when having many zstd processes at once.