[DOCS-11288] Recommend allocating separate device for the dbpath/journal Created: 06/Feb/18 Updated: 30/Oct/23 |
|
| Status: | Closed |
| Project: | Documentation |
| Component/s: | manual, production |
| Affects Version/s: | None |
| Fix Version/s: | Server_Docs_20231030 |
| Type: | Improvement | Priority: | Major - P3 |
| Reporter: | Shakir Sadikali | Assignee: | Unassigned |
| Resolution: | Won't Do | Votes: | 5 |
| Labels: | production-notes | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Participants: | |
| Days since reply: | 1 year, 14 weeks, 2 days ago |
| Epic Link: | DOCSP-1769 |
| Description |
|
We have encountered numerous instances of performance issues resulting from a collision of io requests from checkpoints and journal writes. The symptoms are a known performance issue resulting from placing the dbPath directory and the WiredTiger journal directory on the same filesystem. The bursts of IO to WiredTiger datafiles that occur at the end of each checkpoint when filesystem buffers are flushed to disk can saturate the disk, delaying writes to the journal files, and producing small IO "stalls" exactly like those observed on the mongod logfiles. The remedy to this problem is relatively simple – to place the mongod dbPath directory and the mongod dbPath/journal directory on separate physical devices and use a symlink to point to the journal directory from the mongod dbPath. Physically separating the IO removes the contention between database files and journal files during checkpoints, and prevents the small IO stalls observed in this case and others like it. |
| Comments |
| Comment by Education Bot [ 31/Oct/22 ] |
|
Hello! This ticket has been closed due to inactivity. If you believe this ticket is still important, please reopen it and leave a comment to explain why. Thank you! |
| Comment by Kelsey Schubert [ 06/Feb/18 ] |
|
I think the important distinction that needs to be clarified is that even if disk capacity is not regularly fully utilized, users may see improved query latency by moving the journal to a different path. |
| Comment by Shakir Sadikali [ 06/Feb/18 ] |
|
Hi ravind.kumar - I have. I believe we need to expand on that description. Perhaps use some of the text in the description (written by mark.brinsmead) to elaborate on the rationale for the recommendation? |
| Comment by Ravind Kumar (Inactive) [ 06/Feb/18 ] |
|
Hi shakir.sadikali - have you seen this section in our production notes?. I believe it covers your concerns. |