[SERVER-53599] Log collection name on invariant. Created: 06/Jan/21 Updated: 29/Oct/23 Resolved: 26/Feb/21 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | None |
| Affects Version/s: | None |
| Fix Version/s: | 5.0.0 |
| Type: | Improvement | Priority: | Major - P3 |
| Reporter: | Luke Pearson | Assignee: | Bynn Lee |
| Resolution: | Fixed | Votes: | 1 |
| Labels: | newgrad | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Backwards Compatibility: | Fully Compatible |
| Sprint: | Execution Team 2021-03-08 |
| Participants: |
| Description |
|
While working on the WT_NOTFOUND issue encountered by a number of users recently we've been looking at one particular invariant:
This invariant has proven to be very useful however on a number of occasions we've wanted to figure out which collection was being operated on when this invariant was hit (and a few other bits of info). As such I'd like this invariant and perhaps other to log more information. For this one in particular I'd say the following pieces of information would be helpful:
Open for feedback. |
| Comments |
| Comment by Githook User [ 26/Feb/21 ] |
|
Author: {'name': 'Bynn Lee', 'email': 'bynn.lee@mongodb.com', 'username': 'bynn'}Message: |
| Comment by Geert Bosch [ 14/Jan/21 ] |
|
You're right it's still here. That seems problematic in a way, as collection names can change, but we still need to be able to have concurrent access to the RecordStore and there is no mutex protecting this. Also the question on what the name of the collection associated with a recordstore is (if it's associated with a collection at all) depends on the timestamp. It was quite a bit of trouble for the lock-free reads project to remove the _ns objects all around,, and there were also bugs related to missing updates. |
| Comment by Luke Pearson [ 11/Jan/21 ] |
|
For simplicity then, I'll update the ticket to just be logging of the collection name. While the read timestamp and key could be useful it feels use case specific. |
| Comment by Louis Williams [ 08/Jan/21 ] |
|
The collection name is available as _ns. |
| Comment by Luke Pearson [ 07/Jan/21 ] |
|
I thought that might be the case, would that then be the UUID of the collection? Can that be used to get the customers collection name eventually? |