CBT: the thread to centralize your feedback
-
@olivierlambert Looks like it's back to single threaded bottlenecks...
I see a lot of single core 100% utilization on the XO VM.
-
@Andrew Hi Andrew, can't reproduce on my end, all cores utilized at the same time around 30 to 40 % for 2 simultanious backups.
-
@rtjdamen It happens when Continuous Replication is running. The source and destination and network can do 10Gb/sec.
I'll have to work on a better set of conditions and tests to replicate the issue.
I know it's slower because the hourly replication was taking 5-6 minutes and now takes 9-10 minutes. It's more of an issue when the transfer is >300GB.
Just feedback....
-
@Andrew understood! We do not use that at this time.
-
@olivierlambert Hi Olivier, do you have an ETA?
-
Today
-
@olivierlambert Running current XO master (commit 1ace3), using Backup and running a full backup to a NFS remote, it now ignores the
[NOBAK]on a disk.Also, if you cancel the export task the backup is still marked as Successful. I never tried that before so it may have always done that.
-
@Andrew Thanks for your feedback, we are currently investigating this issue, we have difficulties reproducing this issue though.
Also, if you cancel the export task the backup is still marked as Successful. I never tried that before so it may have always done that.
Yes, this is not supported, the task is only here for informational purposes.
-
@Andrew i tried the same on our end, changed one of the disk names to [NOBAK] in front but the disk is still processed. Not a big of a deal for us, we do not use this currenlty but wanted to confirm we see the same behavior.
I also tested with a new vm, created 2 disks, added [NOBAK] in front of the diskname, ran backup job and both disks are exported.
@olivierlambert are u still able to deliver the update to latest today?
-
@julien-f Updated to Xen Orchestra, commit 1ace3
Running my delta backups now all worked without issues. Also Coalesces finished fine.
The only thing I note is that there is always one (or +1) tasks showb when there is actually none running:
-
@manilx It may be a normal task that's hidden (like SR or ISO scan). Clear out the search box by just clicking on the X. It should show all tasks.
-
@rtjdamen Yes
-
@olivierlambert just updated, we will run some jobs with snapshot destroy and will report the results
Compliments to all involved fot this great improvement is such a short time!
-
How do I get rid of this one?
-
@manilx restart toolstack did it. Thx for pointing this out
-
@rtjdamen That was a big sprint
-
@olivierlambert indeed but the first version is here!
I allready noticed some strange behavior, when a backup job is running with both switches on (cbt and data destroy), the first run the job processes succesfull but some snapshots stay orphan after the job is finished and coalesced. @florent any idea what is causing this?
-
@olivierlambert With today's update (feat: release 5.96.0, master 96b7673) things are working well again. I have not seen any Orphan VDI/snapshots or stuck Control Domain VDIs (but it may take time to test). Thanks to the Vates team!
-
@Andrew need to do some more testing but second run does clean them out, keep u all posted on the results over the weekend!
-
I did some digging on this orphan snapshots. i found this error in the backup logging for these vms
Couldn't deleted snapshot data
error
{"code":"VDI_IN_USE","params":["OpaqueRef:d52c23f7-4008-45e0-9852-6ecb3aeb8567","data_destroy"],"call":{"method":"VDI.data_destroy","params":["OpaqueRef:d52c23f7-4008-45e0-9852-6ecb3aeb8567"]}}
vdiRef
"OpaqueRef:d52c23f7-4008-45e0-9852-6ecb3aeb8567"i do not see them currently on normal vms and also not on second run
