[email protected] wrote: > Quoting Ben Rockwood <[email protected]>: > >> [email protected] wrote: >>> I noticed alot of this where the pool members had no activity >>> according to iostat, but the ssd being used heavily. I don't really >>> understand, but can someone give me some insight into this? >> >> The SSD doing SLOG (ZIL) is recording all the synchronous writes as they >> come in and then returning the "ok" to the calling process. Meanwhile, >> the write transactions are queuing up in memory (a "transaction group", >> txg) waiting to be flushed to physical disk (a "txg sync") every 30 >> seconds. >> >> If you watch iostat for a while, you should see almost constant activity >> to the SLOG disk and then a big burst of write IO approx every 30 >> seconds; possibly more frequently depending on how fast your writing >> data into memory. >> >> >> benr. >> > > Thanks very much again. > What happens if the log device for a pool dies suddenly?
If the log device dies it resorts to writing ZIL to the normal pool disks, as though you never added one. The only downside is that in that event the performance will substantially slow down untill its repaired, which is why you have the option to use mirrored SLOGs. benr. _______________________________________________ storage-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/storage-discuss
