Uhm, is ARPNetworks having an outage?
Nevermind.
loudnoises!
shhhhhh qbit no news is good news, it means that everything is working swimmingly and we're all glad that up_the_irons is on the west coast
also glad my VPS is west coast
mikeputnam: yea…until the big one hits
doing a "portmaster -Da" on a server recently transferred (via ZFS) from a VPS to one of our dedicateds.  my gawd this thing FLIES.
i mean.. it's disgustingly fast
lol
up_the_irons: non-shared disk will do that ;)
jdoe: but it's not just that... i mean, this thing is so fast i need a seatbelt
jdoe: i wonder if ZFS is caching a lot in RAM.. it has 32GB
I dunno man. For me, the bottleneck with ports stuff (even on dedicated hardware) was *always* disk
could be, it caches aggressively.
buy some ssd's and set up as caches on your zpool's
gives you a very nice performance boost
seems kinda sketchy, given the failure rate of ssds.
wonder if the checksums will save you from an unreliable l2arc
jdoe: jbergstroem : i have a client that does that (ssd caching), he's reported great success
the performance i've just seen though, seems to be quite good enough
the disks aren't even fancy, just 7200RPM SATAs
2x (mirror)
same... but then, I only used zfs for personal stuff.
I do wonder though, what happens if your l2arc fails.
l2arc?
the ssd (or whatever) cache.
arc (memory) -> l2arc (ssd) -> disk
according to:
http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide#Storing_ZFS_Snapshot_Streams_.28zfs_send.2Freceive.29
"If a cache device fails, the data will simply be read from the main pool storage devices instead"
(better link: http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide#Separate_Cache_Devices)
up_the_irons: it's used as a write cache too though.
jdoe: i would imagine that, since everything written is checksumed, if an error were to occur, it may report a write error up the chain or attempt to bypass cache
yeah... unfortunately that's the sort of thing I'd want to test before it happened.
haha.
... and I have no idea how.
though it's probably safe to assume smarter people than me ensured that it does ;)
jdoe: it'd be pretty easy to test, just pull out the SSD during some write intense operations and see if the server keeps going :)
haha.
right. I also don't have backups.
heh
i can try that tomorrow
i can try while setting it to offline as well
well. that was boring
write speeds went down, nothing else