istgt requires a restart everytime LUNs are changed, which is very cruel. Also any connectivity issue very easily causes client node to remount as read only, for unknown reason. Therefore, we are possibly changing the iSCSI daemon to another one within the next 48hrs which was noted to handle errors much more gracefully in tests done ... Read More »
Today after the maintenance we've noticed bad client side performance - unlike before when performance was fine and good. The file system itself seems to be going really fast right now, we are seeing 1.6G/s (YES, Gigabytes per second) writes and reads, with random reads still above 540M etc. with iozone -a -s 16g -r 4096 -T test. This needs more ... Read More »
Storage maintenance took much longer than expected, but the configuration error has been fixed now and performance should be increased very significantly.
All vnodes are back online, and we are prepping to bring online a lot more during this weekend.
We need to do some storage system maintenance to gain even more performance out of it.
This will happen today, approximately from 18:00 GMT, and will take couple hours.
We will try to keep client nodes up & running during this period but we cannot guarantee that. Downtime is likely to occur.
