Endless Xapi#getResource /rrd_updates in tasks list
-
Thanks, might be useful for @julien-f
-
@olivierlambert
Same on my host, the first stuck task appeared 5 hours after toolstack and XO reboot. -
8b7e1 still got the tasks
-
-
@uwood 0794a still some tasks
-
That's tricky, @julien-f couldn't reproduce it in our lab. So we'll need more details on your setup guys.
-
@olivierlambert
I have one pool, and three Nodes in my pool.
Node 1 – Dell R420, 2x 10gb NICs (Master) – 192.168.1.4
Node 2 – Dell R620, 2x 10gb NICs – 192.168.1.7
Node 3 – Dell R420, 2x 10gb NICs – 192.168.1.3
XO VM – Ubuntu 22.04.4 LTS – 192.168.1.15Each Node has a Bond of NIC 2 + 3. (Node 2 with the R620 has the mac addresses re-assigned to work correctly)
Above each node, network wise, is 2x switches (2x Unifi USW-Agg) and I use the Unifi Dream Machine Pro as my router. I am able to ping the other Nodes from each Node.
Local DNS utilizes Technitium DNS (Primary & Secondary), as a recursive DNS.
My SR’s are two iSCSI datastores that run on a separate server running TrueNAS Scale on a Dell R320.
Within my pools, I run about 25x virtual machines. I run nightly backups for ~5 VM’s, and weekly backup’s for all VM’s. Backups have a remote NFS storage location hosted on a separate server. I have 3 VM’s that run on separate network vlans than the rest, and those networks are setup under the pool, and upstream on the router.
Plugins, I have the following enabled
• Backup-reports
• Load-balancer (performance mode)
• Perf-alert
• Transport-email
• Usage-reportFrom my testing, this was introduced with commit 6c16055 - Mar 15. I have since rolled back to c6451cf and have stayed on this commit for the past several days.
-
Thanks for the details @14wkinnersley !
-
@14wkinnersley can you disable perf alert and load balancer plugin and see if it still happening?
-
@olivierlambert
Simpler setup on my side:Pool 1 - 1 x ASRock Rack 1U4LW-X570/2L2T RPSU with about 10 VMs
Pool 2 - 1 x HP DL360 Gen9 with only the XO VM (Debian 11).Each host has 2 x 10 Gb NIC in use (one for VMs/MGMT, the other one for NFS storage).
The storage is a NFS share on a QNAP NAS (except for the XO VM which is on the local storage of the DL360 host).
The 2 hosts and the NFS storage are connected on the same10Gbit Ubiquiti EdgeSwitch.
Same plugins as @14wkinnersley + audit and sdn-controller.
The issue with stuck task is only appearing on the pool 1, not on the pool 2 with only the XO VM.
-
@olivierlambert Will do. Plugins are disabled and I'm updating back to master right now. Will report back.
-
Thank you very much, both of you
-
I'm gonna try the same and will let you know.
-
@olivierlambert Updated to master (Commit: 79c9ef0) with perf alert and load balancer plugin's disabled. Cleared the toolstack for good measure, and I got no hanging tasks.
For some further testing, I flipped the load balancer plugin back on and waited about 5 hours and received no hanging tasks. I turned load balancer plugin back off (Just in case), cleared the toolstack, and turned the perf alert plugin on and received the first hanging task within about 10 minutes, and subsequently had more tasks hang every ~10-12 minutes.
-
That would have been my next question
So it's the perf alert thing likely. Thank you, this will help @julien-f a lot -
@olivierlambert
It took a while to validate it, but I can also confirm that the perf-alert plugin is the culprit.Here are my plugin's parameters, if it can help chasing the goose:
-
@Mathieu I just activated this plugin as a test on our 2 EPYC hosts and I also get these tasks now. So yes, this plugin has an issue.
Deactivated.
-
Thanks @julien-f is now reviewing the plugin code
-
I am seeing this same issue on master build cc049 if I turn on perf-alert it starts with in a few minutes. If I turn it off and restart the tool stack it stops.
-
It's been fixed since a bit now. Double check you have the correct plugin code rebuilt.
