Good news: It's look like 100% wait stalls bug appearing on si3124 driver (and thus on si3132 chipsets which use that driver) is fixed in recent nevada builds. I replaced binary si3124 driver with one from snv_145 provided by Illumos IPS repo for upgrading Opensolaris b134 build os in order to compile Illumos (see their build tutorial at: http://www.illumos.org/projects/illumos-gate/wiki/How_To_Build_illumos p. 3 'Upgrade to b145'). My subjective feeling is that fix for Bug ID 6786704 (fixed in snv_136) also fixed this bug.
I tested my si3132 based storage throughly for few weeks and I didn't noticed write stall nor performance degradation that I experienced (and described in this topic) earlier. I found only one issue with 100% wait bug on snv_145 si3124 driver - 100% wait bug may appear occasionally when zpool import takes place - and in that case it least forever completly blocking i/o on imported zpool and thus leading to ZFS hangup (it's impossible to execute another zfs cmd). -- This message posted from opensolaris.org _______________________________________________ storage-discuss mailing list storage-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/storage-discuss
