Hi, i've stumbled upon an issue with the following change to -current about 8 weeks ago:
/sys/ufs/ffs/ffs_softdep.c r1.146 https://marc.info/?l=openbsd-cvs&m=157815505920104&w=2 All systems (softdep enabled) were running -current dated 23 Dec 2019 and after updating to -current 13 Jan 2020, one system hosting a fairly busy postgres database became unresponsive and started to eat up 100% cpu...about 30m - 6h after the last boot. The system runs as VM on ESXi 6.7u3 (AMD Epyc gen1) and unfortunately... even after several weeks of testing it didn't crash a single time, so no ddb, no entries in /var/log...the system is simply "busy" and won't react to anything (hypervisor monitoring shows that as soon as the system becomes unresponsive, cpu utilization goes up to 100%). I've tested with another VM running on ESXi 6.7u3 (AMD Epyc gen2)...same results. In order to avoid any hardware related issue, i've setup the exact same OpenBSD-current + configuration on a bare metal Intel platform (Supermicro X10SLM-f Xeon E3-1230v3) with a serial cable connected. A few hours after boot it showed the identical error. (i did several tests over two weeks). Unfortunately, even with ddb.panic=1 ddb.console=1 and sending a BREAK with #~ (cu -l /dev/cuaU0 -115200) i don't get any helpful info (no trace or ps). ~# The following connections are open: #1 client-session (t4 r0 i0/0 o0/0 e[write]/0 fd 5/6/7 sock -1 cc -1) #2 mux-control (t16 r3 i0/0 o0/0 e[closed]/0 fd 8/8/-1 sock 8 cc -1) #3 client-session (t4 r1 i0/0 o0/0 e[write]/0 fd 9/10/11 sock -1 cc 2) #4 mux-control (t16 r5 i0/0 o0/0 e[closed]/0 fd 12/12/-1 sock 12 cc -1) #5 client-session (t4 r2 i0/0 o0/0 e[write]/4 fd 13/14/15 sock -1 cc 4) It won't react to commands. So as mentioned in the beginning, when running a fresh -current kernel but w/o the change in /sys/ufs/ffs/ffs_softdep.c r1.146, the system is stable, even after weeks...and...the same applies for an unmodified -current kernel, but with mounting /var/postgresql w/o softdep. Thanks in advance for any info - especially to get more debugging output. -Mark -- Mark Patruck ( mark at wrapped.cx ) GPG key 0xF2865E51 / 187F F6D3 EE04 1DCE 1C74 F644 0D3C F66F F286 5E51 https://www.wrapped.cx
