We recently setup two identical RH61 servers (mail and POP servers). We configured quotas using linuxconf, one system we left in place, the other system had to be powered off and moved to a different rack. Both systems have the same number of users (we have an administration system that balances users as they are added/removed). They have similar load and each has ~5000 users each with a home directory and a mail spool quota. On the console of the non-rebooted system, the kernel slowly printed these messages (since they didn't seem harmful we just left it alone. However, as you read down, there's a VFS file-max limit warning and after that it was impossible for the system to spawn new processes. Oct 29 20:30:06 grey kernel: get_empty_dquot: pruning 376 Oct 29 20:41:52 grey kernel: get_empty_dquot: pruning 488 Oct 29 20:49:46 grey kernel: get_empty_dquot: pruning 414 Oct 29 21:00:00 grey kernel: get_empty_dquot: pruning 478 Oct 29 21:19:57 grey kernel: get_empty_dquot: pruning 381 Oct 29 22:37:42 grey kernel: get_empty_dquot: pruning 524 Oct 29 23:09:59 grey kernel: get_empty_dquot: pruning 404 Oct 29 23:29:47 grey kernel: get_empty_dquot: pruning 399 Oct 30 02:07:27 grey kernel: get_empty_dquot: pruning 132 Oct 30 12:19:33 grey kernel: get_empty_dquot: pruning 128 Oct 30 12:19:47 grey kernel: get_empty_dquot: pruning 128 Oct 30 12:19:47 grey kernel: get_empty_dquot: pruning 565 Oct 30 12:37:50 grey kernel: get_empty_dquot: pruning 556 Oct 30 12:47:41 grey kernel: get_empty_dquot: pruning 337 Oct 30 12:58:21 grey kernel: get_empty_dquot: pruning 436 Oct 30 13:07:38 grey kernel: get_empty_dquot: pruning 481 Oct 30 13:29:01 grey kernel: get_empty_dquot: pruning 790 Oct 30 14:16:15 grey kernel: get_empty_dquot: pruning 507 Oct 30 14:47:58 grey kernel: VFS: file-max limit 4096 reached Oct 30 14:49:00 grey kernel: Unable to load interpreter Oct 30 14:54:44 grey kernel: Unable to load interpreter Oct 30 14:59:04 grey kernel: Unable to load interpreter The rebooted system is still working fine.
Assigned to dledford
6.1 certainly had problems with quotas. If the 6.2 erratakernel shows the same ever, then re-open the bug.