Hi

Note that the plugin:zfs-send instance was put into maintenance at the
request of time-slider:default - ie it used something likens cadmium mark.

Looking at the start method for time-slider (vs prop -p start/exec
time-slider) function check_failure performs svcadm mark maintenance for
failed commands.  The only use of
check_failure on my s12 system is after frontal edit, but yours may vary.

Worst case I would add some strategic set -x to the start method script and
then have a look at the svc log file after failure to see where it goes
wrong.

Also worth knowing that time-slider has a "feature" whereby it only manages
leaf datasets, ie those with no child datasets.  A real pest. Likely easy
to fix, but it is not seen as an enterprise feature nowadays, I think.

Gavin

On Thursday, 21 November 2013, Chris Wells wrote:

> OK, I've now completely disabled timeslider, deleted all snapshots,
> deleted all BEs, other than the default, reneabled timeslider, cleared
> any svc issue, and ensured that snapshots are enabled:
>
> # zfs set -r snapdir=visible rpool
> # zfs set -r com.sun:auto-snapshot=true rpool
> # zfs set com.sun:auto-snapshot=false rpool/swap
> # zfs set com.sun:auto-snapshot=false rpool/dump
>
> SO -
>
> 1) svcs -x still shows svc in maintenance mode:
>
> # svcs -x
> svc:/application/time-slider/plugin:zfs-send (Time Slider Plugin Framework)
>  State: maintenance since November 21, 2013 01:18:21 AM EST
> Reason: Maintenance requested by "svc:/application/time-slider:default"
>    See: /var/svc/log/application-time-slider:default.log
>    See: http://support.oracle.com/msg/SMF-8000-R4
>    See: /var/svc/log/application-time-slider-plugin:zfs-send.log
> Impact: This service is not running.
>
>
> 2) snapshots are missing for / (rpool/ROOT/SRU11.1.12.5.0) :
>
> # df -k /
> Filesystem           1K-blocks      Used Available Use% Mounted on
> rpool/ROOT/SRU11.1.12.5.0
>                      398118676   4909286 393209390   2% /
> # df -k /.zfs
> Filesystem           1K-blocks      Used Available Use% Mounted on
> rpool/ROOT/SRU11.1.12.5.0
>                      398118737   4909286 393209451   2% /
> # ls /.zfs
> shares  snapshot
> # ls /.zfs/snapshot
> #
>
>
> Any ideas? Anyone?
>
> ---------- Forwarded message ----------
> From: Chris Wells <[email protected] <javascript:;>>
> Date: Tue, Nov 12, 2013 at 8:35 AM
> Subject: Solaris 11.1 Timeslider
> To: [email protected] <javascript:;>
>
>
> Hi guys.
>
> I've been having a problem with Timeslider (well, since about
> OpenSolaris Nevada 130 something).
> Having done a complete reinstall of Solaris 11.1, and brought it up to
> the current SRU (SRU11.1.12.5.0), I'm still having issues.
>
> What I'm finding is that timeslider fills up past the autodelete level
> (75%), and generally gets bogged down, and fills up to 100% eventually
> without corrective action.
>
> And the svc:/application/time-slider/plugin:zfs-send service (Time
> Slider Plugin Framework), usually gets put into maintenance mode by
> time-slider.
>
> What I've found is that time-slider puts zfs holds (why?) on zfs
> snapshots, and the snapshots never get deleted.
>
> I can unblock it with something like:
>
> fixtimeslider.sh :
> ------------------------
> #!/bin/sh
>
> svcadm disable -t time-slider
>
> for f in `zfs list -H -t snapshot -o name`; do \
>   zfs release -r org.opensolaris:time-slider-plugin:zfs-send $f;
> done
>
> svcadm clear svc:/application/time-slider/plugin:zfs-send
>
> svcadm enable time-slider
> ------------------------
>
> I ran this over the weekend, and was able to reduce the zpool
> utilisation down from 95% used down to 60%. However, zfs held
> snapshots have started appearing again:
>
> root@marvin:~# for f in `zfs list -t snapshot -o name`; do zfs holds
> $f; done 2>&1 | grep "zfs-send"
> rpool@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:50 2013
> rpool/ROOT/SRU11.1.11.4.0/var@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:49 2013
> rpool/ROOT/SRU11.1.12.5.0/var@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:50 2013
> rpool/ROOT/s11.1ga/var@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:48 2013
> rpool/VARSHARE@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:49 2013
> rpool/VMs@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:49 2013
> rpool/download@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:47 2013
> rpool/export/home/crispi@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:47 2013
> rpool/export/users@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:48 2013
> rpool/ftp@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:46 2013
> rpool/incoming@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:46 2013
> rpool/isos@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:47 2013
> rpool/squid@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:45 2013
> rpool/www@zfs-auto-snap_hourly-2013-11-11-00h46
> org.opensolaris:time-slider-plugin:zfs-send  Mon Nov 11 00:46:45 2013
>
>
> So - Why is this happening? And what is the long term fix? - surely
> zfs/timeslider should be more reliable than this?
>
>
> --
> Regards,
>
> Chris
>
>
> --
> Regards,
>
> Chris
>
> _______________________________________________
> msosug mailing list
> [email protected] <javascript:;>
> http://mexico.purplecow.org/m/listinfo/msosug
>
_______________________________________________
msosug mailing list
[email protected]
http://mexico.purplecow.org/m/listinfo/msosug

Reply via email to