[SERVER-17098] wiredTiger read checksum error Created: 28/Jan/15 Updated: 11/Mar/15 Resolved: 24/Feb/15 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Storage |
| Affects Version/s: | 3.0.0-rc7 |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Critical - P2 |
| Reporter: | Quentin Conner | Assignee: | Michael Cahill (Inactive) |
| Resolution: | Cannot Reproduce | Votes: | 0 |
| Labels: | 28qa | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Environment: |
CentOS 6.5 |
||
| Attachments: |
|
| Operating System: | Linux |
| Steps To Reproduce: | Run sysbench against mongod on the same host (CentOS 6.5) in a single-node topology with Java driver ver 2.12.4. |
| Participants: |
| Description |
|
With single-node sysbench workload running insert, mongod 3.0.0-rc7 from MCI encountered a wiredTiger library panic.
|
| Comments |
| Comment by Michael Cahill (Inactive) [ 24/Feb/15 ] |
|
The original issue cannot be reproduced. Please open a new ticket if it happens again. quentin.conner, we have some open tickets around memory and pauses: if you can reproduce them in RC9 and above and there are no matching tickets, please open new ones. |
| Comment by Quentin Conner [ 10/Feb/15 ] |
|
Could not reproduce the WT panic upon re-execution of the same automated test immediately after reboot. |
| Comment by Quentin Conner [ 05/Feb/15 ] |
|
The same sysbench workload was run against the same machine successfully (data load), then run in execute mode for six days without error. This is on a rotating magnetic hard disk. Not able to reproduce this symptom yet. |
| Comment by Michael Cahill (Inactive) [ 02/Feb/15 ] |
|
Let me know if you can reproduce this: at the moment I don't have enough information to make progress. Just for clarity, this kind of error would be generated if files in the database directory were modified by some other process while mongod is running. For example, if a backup was restored over the database directory while mongod was performing updates, an error like this could be generated. To my knowledge, we haven't seen this error from any of the workloads that we have run against mongod without some kind of external corruption, and sysbench isn't doing anything much different to the tests we're running continuously. |
| Comment by Quentin Conner [ 29/Jan/15 ] |
|
Nothing noteworthy. Fresh boot of the machine. EXT4 filesystem that had been in use for RC6 testing previously. It was a clean database in that it did not exist (no files on disk) at the begin of the run. It will be a few days to see if it reproduces. The machine is busy with another workload. |
| Comment by Michael Cahill (Inactive) [ 29/Jan/15 ] |
|
Was this with a clean database at the beginning of the run? Is it reproducible? To my knowledge, we haven't seen anything like this and it is a frequent code path in WiredTiger. Is there anything unusual about the host (filesystem, storage subsystem, etc.)? |
| Comment by Quentin Conner [ 28/Jan/15 ] |
|
mongod log file attached as mongodb-sysbench.log |