Details
-
Bug
-
Resolution: Fixed
-
Critical
-
None
-
Lustre 2.4.0-19chaos
-
3
-
11907
Description
We are using lustre 2.4.0-19chaos on our servers running with the ZFS OSD. On some of the OSS nodes we are seeing messages like this:
Nov 6 00:06:29 stout8 kernel: LustreError: 14909:0:(osd_object.c:973:osd_attr_set()) fsrzb-OST0007: failed to update accounting ZAP for user 132245 (-2) Nov 6 00:06:29 stout8 kernel: LustreError: 14909:0:(osd_object.c:973:osd_attr_set()) Skipped 5 previous similar messages Nov 6 00:06:38 stout16 kernel: LustreError: 15266:0:(osd_object.c:973:osd_attr_set()) fsrzb-OST000f: failed to update accounting ZAP for user 122392 (-2) Nov 6 00:06:38 stout16 kernel: LustreError: 15266:0:(osd_object.c:973:osd_attr_set()) Skipped 3 previous similar messages Nov 6 00:06:40 stout12 kernel: LustreError: 15801:0:(osd_object.c:973:osd_attr_set()) fsrzb-OST000b: failed to update accounting ZAP for user 122708 (-2) Nov 6 00:06:40 stout12 kernel: LustreError: 15801:0:(osd_object.c:973:osd_attr_set()) Skipped 4 previous similar messages
Nov 7 00:31:36 porter31 kernel: LustreError: 7704:0:(osd_object.c:973:osd_attr_set()) lse-OST001f: failed to update accounting ZAP for user 54916 (-2) Nov 7 02:53:05 porter19 kernel: LustreError: 9380:0:(osd_object.c:973:osd_attr_set()) lse-OST0013: failed to update accounting ZAP for user 7230 (-2)
Dec 3 12:01:21 stout7 kernel: Lustre: Skipped 3 previous similar messages Dec 3 13:52:30 stout4 kernel: LustreError: 15806:0:(osd_object.c:967:osd_attr_set()) fsrzb-OST0003: failed to update accounting ZAP for user 1752876224 (-2) Dec 3 13:52:30 stout4 kernel: LustreError: 15806:0:(osd_object.c:967:osd_attr_set()) Skipped 3 previous similar messages Dec 3 13:52:30 stout1 kernel: LustreError: 15324:0:(osd_object.c:967:osd_attr_set()) fsrzb-OST0000: failed to update accounting ZAP for user 1752876224 (-2) Dec 3 13:52:30 stout1 kernel: LustreError: 15784:0:(osd_object.c:967:osd_attr_set()) fsrzb-OST0000: failed to update accounting ZAP for user 1752876224 (-2) Dec 3 13:52:30 stout14 kernel: LustreError: 16345:0:(osd_object.c:967:osd_attr_set()) fsrzb-OST000d: failed to update accounting ZAP for user 1752876224 (-2) Dec 3 13:52:30 stout12 kernel: LustreError: 32355:0:(osd_object.c:967:osd_attr_set()) fsrzb-OST000b: failed to update accounting ZAP for user 1752876224 (-2) Dec 3 13:52:30 stout2 kernel: LustreError: 15145:0:(osd_object.c:967:osd_attr_set()) fsrzb-OST0001: failed to update accounting ZAP for user 1752876224 (-2) Dec 3 13:52:30 stout10 kernel: LustreError: 14570:0:(osd_object.c:967:osd_attr_set()) fsrzb-OST0009: failed to update accounting ZAP for user 1752876224 (-2)
First of all, these messages are terrible. If you look at osd_attr_set() there are four exactly identical messages that are printed. Ok, granted, we can look them up by line number. But even better would be to make them unique.
So looking them up by line numbers 967 and 973, it would appear that we have hit at least the first two of the "filed to update accounting ZAP for user" messages.
Note that the UID numbers do not look correct to me. Many of them are clearly not in the valid UID range. But then I don't completely understand what is going on here yet.
Attachments
Issue Links
Activity
Ahh OK. I remember now.
It looks to me like you're failing to call dmu_tx_commit() after dsl_sync_task_nowait(). The commit is responsible for dropping all the dnode holds and notifying any waiters. Without it the holds are just going to accumulate and I'd expect to see exactly what you're describing.
My suggestion would be to take a good look at the spa_history_log() function in zfs. It's a fairly nice example of how to go about this. In this case they create a tx per history update since there aren't that many of them. In the Lustre case I agree it's probably a good idea to batch them as your doing. However, the commit callback I'd still strongly suggest allowing a dedicated tx for this purpose. I think it would make the code more readable and easy to verify that you've constructed the tx properly. There's a nice comment in include/sys/dmu.h describing what you can and cannot do when constructing a tx. If you break any of those rules you're likely to see some strange problems.
/* * You must create a transaction, then hold the objects which you will * (or might) modify as part of this transaction. Then you must assign * the transaction to a transaction group. Once the transaction has * been assigned, you can modify buffers which belong to held objects as * part of this transaction. You can't modify buffers before the * transaction has been assigned; you can't modify buffers which don't * belong to objects which this transaction holds; you can't hold * objects once the transaction has been assigned. You may hold an * object which you are going to free (with dmu_object_free()), but you * don't have to. * * You can abort the transaction before it has been assigned. * * Note that you may hold buffers (with dmu_buf_hold) at any time, * regardless of transaction state. */
The racy comment was just a subjective feeling I had about the code. There's so much non-private state you're depending on it's hard to easily look at the code and convince yourself it's safe. An good example of this is the reuse the the ot_tx. It's hard to know exactly what state the tx is in when you enter the function, and depending on what state that is there are certain things you must not do. If you were to create a new tx here it would be much clearer.
at umount dnodes storing object accounting are still referenced, so dnode_special_close() gets stuck because meta dnode is referenced by those.
why do you think the whole thing is racy?
Exactly what kind of failure are you seeing? I don't understand what you mean by 'somehow dnodes are still referenced'. Can you point me at a maloo failure which shows the problem or better describe exactly what the issue is. What you're trying to do in the patch looks reasonable to me on the surface, although the whole thing feels racy.
2600 was doing OK on my local system, unfortunately it seem to fail on maloo sometimes. I asked Brian B. to help with understanding the root cause - somehow dnodes are still referenced when I use dsl_sync_task_nowait(). once this sorted out (Brian, please help 
we can try that again.
the important thing is that w/o 2600 accounting is still racy..
Chris,
the patch 10223 was landed for master (2.6.0), which Niu believes to be the major source of inconsistent UID/GIDs on the OSTs for quota accounting.
The 7157 patch is to be tracked under LU-2600 where it was originally filed. I mistakenly thought it was submitted under this ticket and needed a new patch to track it for landing. It had accidentally landed to master for a very short time, but was reverted because it caused problems and Alex had only intended it for testing at this point.
So what is the actual state of the reported bug? Is it now fixed, but because http://review.whamcloud.com/7157 was reverted we now have a potential performance regression? Or is the bug not yet fixed?
The patch http://review.whamcloud.com/7157 was landed to master and then reverted due to problems. That patch needs to be refreshed.
hmm, what I'm trying to do is to hit the same transaction (otherwise accounting is not atomic), but at the point when all the normal changes within this txg are done. - this is the purpose of dsl_sync_task_nowai(), right? and basically that whole state is per txg and it's supposed to be totally non-racy.
yet another point is that this is exactly how space accounting works now? (or very similarly)