[SOLVED] PVESTATD High CPU Usage During MDADM Sync

[Proxygen]

syncing a newly create mdadm raid 1 (WD Red disks, 1.6T partition size, default sync speed limits, internal bitmap enabled) gets CPU load in the 2 to 2.5 range. machine gets sluggish (despite the Xeon E-2136, 6 cores, 12 threads and 32GB RAM)

stopping pvestatd lowers the load to ~1. There is nothing running on this machine yet (no apps, no containers/vms)

what does pvestatd do and is this something that should be looked into?

installed debian 9.9, then pve-manager/5.4-6/aa7856c5 (running kernel: 4.15.18-16-pve)

 

[Alwin]

syncing a newly create mdadm raid 1 (WD Red disks, 1.6T partition size, default sync speed limits, internal bitmap enabled) gets CPU load in the 2 to 2.5 range. machine gets sluggish (despite the Xeon E-2136, 6 cores, 12 threads and 32GB RAM)

I would expect higher load for MDADM, as it syncs up all data and WD Reds are not the fastest spinners around. Aside from that, I want to point out that there is no enterprise support when using MDADM.

what does pvestatd do and is this something that should be looked into?

pvestatd checks amongst others, if the storage is available, what the state of the VM/CTs on that storage are.

 

[Proxygen]

pvestatd checks amongst others, if the storage is available, what the state of the VM/CTs on that storage are.

is it safe to stop pvestatd during mdadm sync? I can live with that.

 

[Alwin]

is it safe to stop pvestatd during mdadm sync? I can live with that.

You may stop the pvestatd, if you can live with no status or stats.