[LU-12021] Error message of mdt_attr_valid_xlate() when 2.12 client to 2.10 server Created: 26/Feb/19 Updated: 27/Nov/19 Resolved: 18/Apr/19 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.11.0, Lustre 2.12.0 |
| Fix Version/s: | Lustre 2.13.0, Lustre 2.12.1 |
| Type: | Bug | Priority: | Minor |
| Reporter: | Li Xi | Assignee: | Qian Yingjin |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Issue Links: |
|
||||||||
| Severity: | 3 | ||||||||
| Rank (Obsolete): | 9223372036854775807 | ||||||||
| Description |
|
A lot of messages of "Unknown attr bits: 0x60000" are printed. And they are very annoying. |
| Comments |
| Comment by Patrick Farrell (Inactive) [ 26/Feb/19 ] |
|
I believe this isn't a client/server interop issue. I looked at this at Cray with 2.11 client & server. I don't have the detailed notes, but I concluded the issue was the kernel/VFS was presenting a new attr bit (ie Linux has added a new attr bit), and the MDS was complaining about it because it didn't recognize it. We decided it was harmless and low priority to fix. We initially saw this only on SLES, but I've seen it on CentOS recently. |
| Comment by Li Xi [ 26/Feb/19 ] |
|
Patrick, I think this is caused by Lazy Size on MDT feature which is just introduced in 2.12. So, not a surprise that 2.11 client doesn't have the problem. |
| Comment by Patrick Farrell (Inactive) [ 26/Feb/19 ] |
|
No, 2.11 did have the problem, lixi_wc. I specifically looked at it and identified the bit as a new attr bit added in newer kernels. I would give more detail but I don't have my notes, since they're internal at Cray. |
| Comment by Li Xi [ 26/Feb/19 ] |
|
Ah, are you still able to get the exact error message? For the flags of 0x60000, I think it is caused by LSoM because these two flags are LSoM flags: #define MDS_ATTR_LSIZE 0x20000ULL /* = 131072 */ I confirmed with the customer that they are using clients of 2.12 with Ubuntu. I guess the 2.11 problem has a different flag value printed in mdt_attr_valid_xlate()? And then it should be caused by some kernel update of the bits. |
| Comment by Patrick Farrell (Inactive) [ 26/Feb/19 ] |
|
I am not, and I just did the same looking as you... I agree, it must've been a different bit, or maybe I'm crazy and we were testing 2.12. I don't think so, but... Certainly, those are the LSOM bits. And there are no new attr bits even in the current Linus kernel... Hm. I will see if I can dig up the error from someone at Cray. Odd. For now, ignore me. |
| Comment by Andreas Dilger [ 27/Feb/19 ] |
|
It would be good to implement one or more of the mechanisms proposed in |
| Comment by Gerrit Updater [ 28/Feb/19 ] |
|
Yingjin Qian (qian@ddn.com) uploaded a new patch: https://review.whamcloud.com/34343 |
| Comment by Gerrit Updater [ 13/Mar/19 ] |
|
Aurelien Degremont (degremoa@amazon.com) uploaded a new patch: https://review.whamcloud.com/34412 |
| Comment by Gerrit Updater [ 15/Apr/19 ] |
|
Minh Diep (mdiep@whamcloud.com) uploaded a new patch: https://review.whamcloud.com/34663 |
| Comment by Gerrit Updater [ 18/Apr/19 ] |
|
Oleg Drokin (green@whamcloud.com) merged in patch https://review.whamcloud.com/34343/ |
| Comment by Peter Jones [ 18/Apr/19 ] |
|
Landed for 2.13 |
| Comment by Gerrit Updater [ 21/Apr/19 ] |
|
Oleg Drokin (green@whamcloud.com) merged in patch https://review.whamcloud.com/34663/ |