pvesh get /nodes/{node}/qemu & qm list very slow since Proxmox 9
- Thread starter Decco1337
- Start date
Any update to this? Issue we have same experience. New cluster on Proxmox 9.1.2 atm, after reading some stuff iam not sure upgrade to 9.1.4 will solve anything. There is currently no easy way redo this on version 8.x
We don't have severe issue as others. With <25 vms per node, but for avg 0,3s to 5,3s is pretty significant. Since our exporters checking `/api2/json/nodes/<node>/qemu` got so much worse and response in exporter raised from 2-5s to 17-25s.
We have KMS disabled
We don't have severe issue as others. With <25 vms per node, but for avg 0,3s to 5,3s is pretty significant. Since our exporters checking `/api2/json/nodes/<node>/qemu` got so much worse and response in exporter raised from 2-5s to 17-25s.
We have KMS disabled
We replicated issue on testing Proxmox 8 after upgrade to 9.1.2, then we upgrade 9.1.4 without success still after while we were around 5s time.
Then we deepdive to smaps which is operation which taking too long:
```
for (; addr != end; pte++, addr += PAGE_SIZE
smaps_pte_entry(pte, addr, walk);
```
So we tried to enable THP always/always. We push agresive converision and after we were 90% done we notice decreased speed from ~5s to ~1,7s.
On production servers we are going now a bit conservative madvise/defer+madvise and see how it goes.
Then we deepdive to smaps which is operation which taking too long:
```
for (; addr != end; pte++, addr += PAGE_SIZE
smaps_pte_entry(pte, addr, walk);
```
So we tried to enable THP always/always. We push agresive converision and after we were 90% done we notice decreased speed from ~5s to ~1,7s.
On production servers we are going now a bit conservative madvise/defer+madvise and see how it goes.
Allright, but thats not in repository as of today. There is still only 9.1.3, so will se when its available.