Hello All,
Got this autosupport from our FAS3020 running 7.2.4. We use the "vol options VOLXX schedsnapname create_time" so that snapshot names reflect when the snapshot was taken, and the autosupport email contains the line:
VOLXX - hourly.2012-05-16_1200 : Snapshot already exists.
Trying to take a snapshot twice? Failing to roll off an old one? Just noise? I've STFW'd a little but I'm not finding anything like this.
Hope to hear from you,
Randy
Hi Randy,
This may be due to how your (NTP) time synchronization is set up. If you sync with a timeserver, this may happen:
- 12:00: the storage system takes snapshot hourly.2012-05-16_1200 on VOLXX - 12:00: the storage system performs a timesync with the NTP server and lears it is actually 11:59:53. The system adjusts the time. - It becomes 12:00 again, the system tries to take the same snapshot again on VOLXX.
This behavior can be avoided by setting the timed.windos option to something like "5m" (5 minutes). The storage system will now not perform a timesync on the hour, but in a timewindow of 5 mins before/after the hour, which greatly reduces this sort of race condition.
Best regards, Filip
On Wed, May 16, 2012 at 9:21 PM, Randy Rue rrue@fhcrc.org wrote:
Hello All,
Got this autosupport from our FAS3020 running 7.2.4. We use the "vol options VOLXX schedsnapname create_time" so that snapshot names reflect when the snapshot was taken, and the autosupport email contains the line:
VOLXX - hourly.2012-05-16_1200 : Snapshot already exists.
Trying to take a snapshot twice? Failing to roll off an old one? Just noise? I've STFW'd a little but I'm not finding anything like this.
Hope to hear from you,
Randy _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
That makes perfect sense. I'll change the NTP setting and we'll see if this happens again...
Thanks!
----- Original Message ----- From: "Filip Sneppe" filip.sneppe@gmail.com To: "Randy Rue" rrue@fhcrc.org Cc: toasters@teaparty.net Sent: Friday, May 18, 2012 2:43:50 AM Subject: Re: Cluster Notification from filer-a (SCHEDULED SNAPSHOT CREATE FAILED) CRITICAL
Hi Randy,
This may be due to how your (NTP) time synchronization is set up. If you sync with a timeserver, this may happen:
- 12:00: the storage system takes snapshot hourly.2012-05-16_1200 on VOLXX - 12:00: the storage system performs a timesync with the NTP server and lears it is actually 11:59:53. The system adjusts the time. - It becomes 12:00 again, the system tries to take the same snapshot again on VOLXX.
This behavior can be avoided by setting the timed.windos option to something like "5m" (5 minutes). The storage system will now not perform a timesync on the hour, but in a timewindow of 5 mins before/after the hour, which greatly reduces this sort of race condition.
Best regards, Filip
On Wed, May 16, 2012 at 9:21 PM, Randy Rue rrue@fhcrc.org wrote:
Hello All,
Got this autosupport from our FAS3020 running 7.2.4. We use the "vol options VOLXX schedsnapname create_time" so that snapshot names reflect when the snapshot was taken, and the autosupport email contains the line:
VOLXX - hourly.2012-05-16_1200 : Snapshot already exists.
Trying to take a snapshot twice? Failing to roll off an old one? Just noise? I've STFW'd a little but I'm not finding anything like this.
Hope to hear from you,
Randy _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
-
Filip Sneppe -
Randy Rue