Reputation
1,265
Top tag
Next privilege 2,000 Rep.
Edit questions and answers
Badges
4 9
Newest
 Yearling
Impact
~63k people reached

Aug
22
comment Is calculating IOPS for ZFS RAIDZ different then calculating IOPS for RAID5 & RAID6?
I just want to be clear - ZFS is by its very design not going to win any performance wars. You can find alternative filesystems & volume managers (or combinations) that can outperform ZFS on the same hardware (ext4 & an Adaptec RAID card, for example). The only exception to this that I'm aware of comes from reads that are easily cached, where the ARC often gives ZFS a leg up on most the alternatives. But when it comes to writes? No. The effort ZFS spends on data integrity and how it handles writes is almost never going to win speed contests. The trade-offs are integrity and feature set.
Aug
22
comment Is calculating IOPS for ZFS RAIDZ different then calculating IOPS for RAID5 & RAID6?
I have a significant amount of experience with this, and can confirm for you that in most situations, RAIDZ is NOT going to outperform the same number of disks thrown into a traditional RAID5/6 equivalent array. Most the traditional RAID5/6 arrays gain IOPS performance (the one almost everyone cares about, even when they don't think they do) as you add spindles to the RAID set, whereas ZFS will not. In return, ZFS won't lose your data, and does not suffer from the 'RAID write hole' problem. Plus that whole snapshots, clones, compression and so on. And the ARC. And .. you get the idea.
Aug
6
comment What happens to missed writes after a zpool clear?
The issue, as I tried badly to explain, is that this is a very 'perfect world' result. It'll happen, but not every time. The reason is ZFS isn't the only thing involved here. FMA, drivers, firmware, lots of other pieces have to behave properly. IF every piece involved properly notices and flags the disk as gone AND as come back, everything will happen without your intervention. That's a very big IF. :)
Aug
6
comment What happens to missed writes after a zpool clear?
'Rather low load' and 'zero load' are two very different scenarios for this circumstance. 'Rather low load' would still cause transaction commits every X seconds, even if it was a single byte of change. Only zero load whatsoever would postpone txg commits (take that statement with a grain of salt; I've never actually tested to see if that's true).
Jul
30
awarded  Revival
Jul
26
awarded  Revival
Jul
25
comment Advice for inserting additional disks to a ZFS mirrored pool?
raidz generally beats mirrors in raw throughput tests, especially at large block sizes. Mirrors beat the everloving piss out of raidz in anything else, especially small block and/or IOPS-constrained tasks. Why no compression for the root filesystem? Nexenta has been compressing the live root filesystem for some time now, as a default.
Jul
25
answered RAID 1E with ZFS
Jul
24
comment partitioning a 2x larger disk to house zfs migration
Are you asking how to consolidate 2 physical vdevs inside a larger pool into 1 physical vdev within that larger pool, or are you asking how to consolidate a zpool made up of just these 2 x 1 TB vdevs into a zpool made up of one 1 x 2 TB vdev? They're very different answers.
Jul
24
comment Multiple 2 TB LUNs — ZFS or ext4-over-LVM
Careful. ZFS' data integrity features are merely unrecoverable errors if the configured zpool is a bunch of LUN's with no ZFS-configured redundancy (eg: mirroring of the LUN's, etc, within ZFS). ZFS will happily tell you (and block a request for) that a block is not matching its checksum and is thus bad -- but it will be unable to correct it if there's no parity (raidz) or other copy (mirror) to go snag (and since ZFS can't communicate with your back-end RAID device, there won't be, from ZFS's perspective). Also, as the responder says -- ZFS performance characteristics will bite you here, too.
Jul
24
comment ZFS: Why does resilvering go over all disks in a pool rather than only those in the raidz1 vdev?
OS? And snipping the zpool status doesn't help here - it'd be useful to see which drives it is saying are resilvering (or does it say none?). It is hard to build an accurate picture from what you've put in here. Usually you'd be correct - a resilver has to rebuild from the rest of a vdev, not the entire pool. However, the 'resilver' message has to give some sort of indicator of progress, and it may be misleading in how it is indicating progress. One thing to look at it would be 'iostat -xen 1' - IS it equally loaded across all drives, or are the ones in that vdev working harder?
Jul
24
comment Why am I getting slow and jumpy ZFS write performance?
Whether or not the ZVOL is actually doing sync writes depends on a number of settings. First, show us the 'sync' setting on the dataset, the writeback cache ('wce' or sometimes 'wcd') setting on the COMSTAR LU, and what if any sync setting is being set by the client? Usually when I see spiky write performance, it is because you're sending too much. ZFS write throttling is a hard nut to crack, lots of moving pieces and tuning to touch. To be frank, it is fairly stupid at times - one of those times is if you slam the box with more than it can handle with no end in sight.
Jul
24
comment Does speed of ZFS snapshot rollback depend on number of files?
That is not technically accurate. A sequential snapshot and a non-sequential file are both completely possible (even probable depending on environment) in ZFS. There is no guarantee that reading through a a file will represent sequential I/O, any more so than a guarantee that reading through an incremental snapshot will represent non-sequential I/O. Probabilities of one being more likely random than the other, certainly, but not guarantees. I'm sorry, but this is just false.
Jul
24
answered What happens to missed writes after a zpool clear?
Jun
1
comment Does speed of ZFS snapshot rollback depend on number of files?
It is, thank you. I've filed what I think is an appropriate answer, based on my own history with zfs send/recv. (something I didn't bother to mention is that the amount of other I/O going on in the pool while doing the zfs send/recv also matters; but the mbuffer solution can actually help mitigate that as well, assuming there's lulls in the non-send/recv I/O).
Jun
1
comment What are the trade-offs between software VPNs (like Hamachi) and dedicated hardware (like SonicWall)?
+1 - My experience with Juniper SRX 2xx's and Cisco ASA 55xx's is good, as well. I'd add you should be careful deploying any hardware VPN device if your environment is one with any BYOD. Especially if they happen to use non-standard OS like Solaris. Even just Mac OSX, sometimes. Most Cisco and Juniper devices, for instance, either do not provide standards-compliant VPN technology (like IPSEC or, ugh, PPTP) - or they do, but only for extra $$. They want you to use proprietary clients and then don't provide those clients for Linux or Mac or Solaris, etc, or those clients cause problems.
Jun
1
answered Does speed of ZFS snapshot rollback depend on number of files?
May
31
comment strange ZFS disk space usage report for a ZVOL
From everything I can see in this, it looks like a bug. The 'used' of a zvol with a volsize of 100G should not exceed much past 100G if there's no children or reservations or the like. Perhaps it was actually a 200+ GB volsize and you changed the volsize parameter? If not, FreeBSD-10.0 is not yet a production release; file a bug with them.
May
31
comment Tuning ZFS scrubbing, 141KB/s running for 15 days
More tuning may have helped, but not necessarily. It is important to note that a ZFS scrub rolls through the data structure, NOT sector by sector on the disks. Which is to say, depending on how the zfs data structure looks on your disks, a scrub operation may look incredibly random -- your disks may be capable of > 100 MB/s of sequential read, but completely random read will be another story entirely. Average block size would also matter here.
May
31
comment Does speed of ZFS snapshot rollback depend on number of files?
This question somewhat confuses me, as you sugest that you transferred a file using netcat @ 410 MB/s, then transferred zfs send/recv over netcat @ 70 MB/s. Then you also meniton ZFS rollback speed. Was the original transfer a zfs send/recv, or not? Where is ZFS rollback coming into play here?