[DOCS-9158] WiredTiger readahead values Created: 18/Oct/16 Updated: 30/Oct/23 Resolved: 27/Jan/17 |
|
| Status: | Closed |
| Project: | Documentation |
| Component/s: | manual |
| Affects Version/s: | None |
| Fix Version/s: | Server_Docs_20231030 |
| Type: | Improvement | Priority: | Major - P3 |
| Reporter: | Luke Prochazka | Assignee: | Ravind Kumar (Inactive) |
| Resolution: | Done | Votes: | 1 |
| Labels: | bite-sized | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Environment: |
Linux |
||
| Issue Links: |
|
||||
| Participants: | |||||
| Days since reply: | 7 years, 2 weeks, 3 days ago | ||||
| Epic Link: | Clean Up ProdNotes | ||||
| Story Points: | 0.5 | ||||
| Description |
|
The current production notes and operations checklist need to have the advice regarding readahead tuning amended for the WiredTiger use case. The new value for WT readahead is 0kb (zero), regardless of media type (spinning disk or SSD), and applies to both the $dbpath and journal directories. |
| Comments |
| Comment by Githook User [ 30/Jan/17 ] |
|
Author: {u'username': u'rkumar-mongo', u'name': u'ravind', u'email': u'ravind.kumar@10gen.com'}Message: Signed-off-by: kay <kay.kim@10gen.com> |
| Comment by Githook User [ 30/Jan/17 ] |
|
Author: {u'username': u'rkumar-mongo', u'name': u'ravind', u'email': u'ravind.kumar@10gen.com'}Message: Signed-off-by: kay <kay.kim@10gen.com> |
| Comment by Ravind Kumar (Inactive) [ 27/Jan/17 ] |
| Comment by Eoin Brazil [ 27/Jan/17 ] |
|
LGTM to me as well. Thanks! |
| Comment by Luke Prochazka [ 24/Jan/17 ] |
|
Hi ravind.kumar, As a generalisation yes, the prefered value is zero, and it applies to all WiredTiger versions. There are always exceptions to the rule though, for instance on platforms that do not support the fadvise() syscall (such as Windows), then FS cache pollution will be a factor. That said, I'm not entirely sure how to word these exceptions. Perhaps something like: "The recommended readahead value for use with the WiredTiger storage engine is 0kb (zero). On supported platforms, WiredTiger utilises the fadvise() syscall to override the Operating System's readahead behaviour and to bypass the filesystem cache. On platforms where fadvise() is not supported (such as Windows), a readahead value of 16 (8kb) has been found to be suitable in some scenarios. These recommendations apply regardless of media type (magnetic or SSD), and applies to both the $dbpath and journal directories." thanks |
| Comment by Ravind Kumar (Inactive) [ 24/Jan/17 ] |
|
Hi all, I'll be taking ownership of this change. luke.prochazka one thing to confirm - is this generally universal for WT, or are there versions of MongoDB where the 0 setting may not be best? |
| Comment by Eoin Brazil [ 24/Jan/17 ] |
|
Is there an outstanding code review you need our input for as I'd like to do whatever I can to help get this published? Thanks! |
| Comment by Alexander Marquardt [ 23/Jan/17 ] |
|
Is this going to be updated in the documentation? I find it more comfortable to make recommendations to the customer when the documentation reflects what I am recommending to them. CC: braz |