Date: Tue, 2 Jan 2018 01:18:59 -0500 From: Garrett Wollman <wollman@bimajority.org> To: freebsd-fs@freebsd.org Subject: Something holding z_teardown_inactive_lock way too long Message-ID: <23115.9299.190414.482629@hergotha.csail.mit.edu>
next in thread | raw e-mail | index | archive | help
I recently upgraded all of our NFS servers to 11.1 after a year of successful operation on 10.3 and (apparently) successful testing of a couple of new servers. But right now I'm seeing that there is something on one of my server (maybe more) that is holding z_teardown_inactive_lock for far, far, far too long, causing alerts to be raised. This server was first booted in 11.1 last Wednesday, and when it's sad, I see this: load: 0.18 cmd: ls 10408 [zfsvfs->z_teardown_inactive_lock] 28.89r 0.00u 0.00s 0% 2576k load: 0.10 cmd: ls 10408 [zfsvfs->z_teardown_inactive_lock] 67.43r 0.00u 0.00s 0% 2576k load: 0.06 cmd: ls 10408 [zfsvfs->z_teardown_inactive_lock] 89.75r 0.00u 0.00s 0% 2576k load: 0.13 cmd: ls 10408 [zfsvfs->z_teardown_inactive_lock] 153.10r 0.00u 0.00s 0% 2576k load: 0.06 cmd: ls 10408 [zfsvfs->z_teardown_inactive_lock] 196.55r 0.00u 0.00s 0% 2576k load: 0.14 cmd: ls 10408 [zfsvfs->z_teardown_inactive_lock] 243.63r 0.00u 0.00s 0% 2576k On the console, it seems like I can still run procstat -kk: 10408 101158 ls - mi_switch+0xe5 sleepq_wait+0x3a _sx_slock_hard+0x334 zfs_freebsd_reclaim+0x3c VOP_RECLAIM_APV+0x89 vgonel+0x21c vnlru_free_locked+0x22c getnewvnode_reserve+0x77 zfs_zget+0x24 zfs_dirent_lookup+0x162 zfs_dirlook+0x77 zfs_lookup+0x44a zfs_freebsd_lookup+0x6d VOP_CACHEDLOOKUP_APV+0x83 vfs_cache_lookup+0xd6 VOP_LOOKUP_APV+0x83 lookup+0x701 namei+0x486 although it did previously hang when I was logging in. Does this sound familiar to anyone? -GAWollman
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?23115.9299.190414.482629>