Reputation
1,345
Top tag
Next privilege 2,000 Rep.
Edit questions and answers
Badges
4 9
Newest
 Yearling
Impact
~75k people reached

Feb
1
comment Is “zfs receive” an atomic operation?
That statement by Matt should be read carefully. The 'reclaim' is similar to a zfs rollback, thus a zfs destroy. If the sent dataset is 10 TB and completed most of it before failing, the resulting reclaim can be incredibly IOPS intensive. Always keep that in the back of your mind when digging into performance oddities, as well as when performing maintenance tasks that involve pool export/import.
Jan
23
comment ZFS: Memory issues with dedup even though zdb -DD looks fine
I thought I posted this but I guess I didn't. Don't assume your data is actually dedupable, btw. Also, in my experience, dedupe ratios usually end up falling over time, not increasing. So if you only start at 3.5x, I wouldn't place bets on it increasing. ZFS dedupe is block level not file level, and not 'intelligent' about bit shift. If you have two otherwise identical blocks of data, but there's a single bit offset on the data, it will not dedupe at all. Most the 'dedupe is what we do' style products tend to have algorithms to look for such things, so they tend to return better ratios.
Jan
22
comment ZFS: Memory issues with dedup even though zdb -DD looks fine
Considering the still-not-insignificant issues regarding 'memory' in the ZFS On Linux project, I would really not recommend you try dedupe at any scale on ZoL. Stick to an illumos-derivative preferably, or a FreeBSD box if that's not an option.
Jan
21
comment XenServer Linux guest booting off iSCSI SAN (ZFS storage): Will a non-checksumming guest file system still be at risk?
Also make sure your full environment is using ECC RAM. The FreeNAS box, and the clients. Non-ECC RAM can silently corrupt data and you'll never know until something crashes/breaks.
Dec
27
comment ZFS: bringing a disk online in an unavailable pool
Very glad you got it to work out, @indotoonster! This is an area that ZFS really needs both more documentation about and an easier/more robust set of options/tools to deal with weird situations. Hopefully developers will donate their time and effort at streamlining this sort of stuff, and it will become better understood as time goes on. :)
Dec
22
answered ZFS: bringing a disk online in an unavailable pool
Dec
11
answered SmartOS using ZFS and guest filesystems
Dec
8
comment How often should I defrag ZFS pools?
I would say not /nearly/ so often. A scrub simply reads every block on the system. That's all it does. By virtue of how ZFS handles reads, ZFS will automatically repair any bad blocks when they're read. This means a scrub is only necessary to read data that isn't commonly being read anyway. Thus, the asker's comment of very busy pools once per week and not so busy pools once per month is actually backwards. A very busy pool, especially one that reads the majority of the live data regularly, need not scrub that often. A quiet pool that is rarely accessed should scrub more often.
Nov
22
comment zfs: zpool space map thrashing - ever fixed?
No, that's not 'solved' on illumos. Or on Linux. Honestly I'm not sure how Oracle could have fixed 'it' (it being a bad, singular-implying word for what is actually a multiple-issue problem). As long as you're comfortable with a high level (not zfs send|recv) method of file transfer, I guess it's OK. Some people have just too much data and too sensitive a downtime requirement for it, or insufficient budget to double up their storage before replacing the OS, etc. :)
Nov
22
comment zfs: zpool space map thrashing - ever fixed?
Well, I focus on Open-ZFS (www.open-zfs.org) 'ports', and my day job is at Nexenta, but you are correct in that I want nothing to do with the beast that is Oracle and its no longer open source code. Given they won't show you the code, I'd be worried that they /have/ done things here, that either through on-disk format change or by virtue of your pool and workload stick you to them, removing your ability to move off Oracle Solaris if you some day need/want to. But enough proselytizing. Good luck. :)
Nov
22
comment size the write cache ZFS
I really need to update that blog post. I forget sometimes to update when I learn new things. :) It isn't actually max speed * txg_commit time. It is max speed * txg_commit * 3. There can actually be up to 3 active transaction groups, one in each of the 3 states of 'active', 'quiescing', and 'syncing'. Thus, for a busy heavy write box, you need a bit more than txg_timeout*3*max write speed in slog devices, assuming all writes are sync.
Nov
12
answered zfs: zpool space map thrashing - ever fixed?
Nov
9
comment Can I remove a vdev from a ZFS pool by rolling back?
Technically? Yes, I suppose so. Much as, technically, I could suddenly have a radiation-induced genetic abnormality take over all my cells and cause me to sprout functional wings that would let me fly myself to the Bahamas. :)
Nov
9
comment Production deployment strategies
Yes - which as long as your app is designed reasonably well and can handle this inconsistency on its side, the only 'defect' would be a different look or functionality for users depending on which box they were hitting. This is not uncommon - Facebook users, for instance, often report seeing functionality or differences that their friends do not. I suspect in Facebook's world, it is doing something to lock users to web heads that have or have not received an update, in some way, as they don't generally report seeing functionality and then suddenly losing it and then gaining it again, etc, etc.
Nov
8
comment Can I remove a vdev from a ZFS pool by rolling back?
Data about the topology of the pool is held in special labels on the leaf vdevs (disks) themselves at the front and end (double copies of each). These 4 entries are redundant copies, not different, and are only 256 KB each in length. They do not have information on history of the layout, merely the layout as it is "now".
Nov
3
answered Can I remove a vdev from a ZFS pool by rolling back?
Nov
1
comment ZFS Recover from Faulted Pool State
Quite concerning that zpool import is just sitting there for hours. That shouldn't happen. Can you try grabbing a live CD or USB bootable image of some fairly new illumos-derivative (OmniOS, OpenIndiana, etc) and see if it can import the pool? The code there is newer than what's in ZoL, AFAIK, so it MIGHT have more success. If not, I suspect your only options are limited to data recovery (very expensive, $100's/hr) or destroy the pool and recreate from backups. Try an OmniOS or OI install/liveCD to import it, first, though.
Nov
1
comment iSCSI: LUNs per target?
To this day the absolute fastest VM performance I've seen with a ZFS backend was Linux Ubuntu clients running KVM with raw files for the disks off an NFSv4 mountpoint from a ZFS NAS head. The perceived performance gain was due in large part to the NFSv4 file delegation mechanics, and the fact that the client machines each had anywhere from 4 to 24 GB of RAM free to throw at local client-side NFS cache (after KVM VM's took their share).
Nov
1
comment iSCSI: LUNs per target?
I rarely see iSCSI beat NFS on performance in VM environments, actually. Well, I see it a LOT, but then I explain COMSTAR's 'writeback cache' (wce) setting, and how it is bypassing the ZIL mechanics of ZFS, and thus not data safe. Then they turn it off, and find iSCSI performs no better or worse than NFS. :)
Oct
31
comment MySQL hangs if connection comes from outside the LAN
That this is going through a VPN and from the tcpdump above, this is very, very likely an MTU-related problem. They are notoriously difficult to diagnose. Force MTU to be identical on client, server, so on, and make sure the max MTU on all switches involved is /higher/ than the the clients (for instance, it's pretty common for max MTU on a switch to be 1528 when the clients are all at 1500). Alternatively you might try forcing the MTU on your VPN down, significantly, try 1400, then 1380, even 1300.