fuse slowdown

Cayuga

Renowned Member
May 3, 2011
86
0
71
After running a three node cluster for years the fuse system /etc/pve dramatically slowed down yesterday (starting vi on a file on /etc/pve can literally take minutes now). The first thing I did was upgrade all three nodes to the latest/greatest software. That didn't help. Can anyone offer suggestions about where to look/what to change to get our old performance back?

Thanks!
 
Doh! I should have looked there first. Yes, there are errors like these:

Nov 12 11:35:17 pie corosync[2618]: [TOTEM ] Retransmit List: 9820
Nov 12 11:35:24 pie corosync[2618]: [TOTEM ] Retransmit List: 983d
Nov 12 11:35:30 pie corosync[2618]: [TOTEM ] Retransmit List: 9859
Nov 12 11:36:33 pie corosync[2618]: [TOTEM ] Retransmit List: 99ad
Nov 12 11:37:21 pie corosync[2618]: [TOTEM ] Retransmit List: 9aa3
Nov 12 11:37:40 pie pvedaemon[3028]: <root@pam> successful auth for user 'root@pam'
Nov 12 11:37:57 pie corosync[2618]: [TOTEM ] Retransmit List: 9b67
Nov 12 11:38:11 pie pmxcfs[2458]: [status] notice: received log
Nov 12 11:39:01 pie corosync[2618]: [TOTEM ] Retransmit List: 9ca6
Nov 12 11:39:30 pie corosync[2618]: [TOTEM ] Retransmit List: 9d4f
Nov 12 11:40:18 pie corosync[2618]: [TOTEM ] Retransmit List: 9e52
Nov 12 11:40:50 pie pveproxy[3054]: worker 251649 finished
Nov 12 11:40:50 pie pveproxy[3054]: starting 1 worker(s)
Nov 12 11:40:50 pie pveproxy[3054]: worker 301342 started
Nov 12 11:41:10 pie corosync[2618]: [TOTEM ] Retransmit List: 9f58
Nov 12 11:41:46 pie corosync[2618]: [TOTEM ] Retransmit List: a01a
Nov 12 11:41:56 pie pmxcfs[2458]: [status] notice: received log

Any idea why they would have started suddenly?

Thanks!
 
(16 processors) 13:25:53 up 1 day, 17:42, 4 users, load average: 1.09, 0.82, 0.93
( 8 processors) 13:25:54 up 3:53, 4 users, load average: 1.93, 1.99, 1.82
(16 processors) 13:25:53 up 1 day, 16:57, 2 users, load average: 1.12, 1.60, 1.64

iperf shows that each of them is passing at least 887 Mbits/sec even with the other traffic on the boxes.
 
Re: fuse slowdown [SOLVED]

It turns out that it was a hardware problem.

Someone swapped a cable; just moved it from one Gigabit switch to another Gigabit switch, but the "new" switch was one which filtered multicast traffic (3COM Baseline 2824).

Thanks for all of your help!