[SERVER-22335] Do not prepare getmore when un-needed in bgsync fetcher Created: 28/Jan/16 Updated: 19/Nov/16 Resolved: 29/Jan/16 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Replication, Sharding |
| Affects Version/s: | None |
| Fix Version/s: | 3.2.3, 3.3.2 |
| Type: | Bug | Priority: | Critical - P2 |
| Reporter: | Timothy Olsen (Inactive) | Assignee: | Scott Hernandez (Inactive) |
| Resolution: | Done | Votes: | 0 |
| Labels: | code-only, csrsupgrade | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Attachments: |
|
||||
| Issue Links: |
|
||||
| Backwards Compatibility: | Fully Compatible | ||||
| Operating System: | ALL | ||||
| Backport Completed: | |||||
| Sprint: | Repl F (01/29/16) | ||||
| Participants: | |||||
| Description |
|
During upgrade from WiredTiger SCCC config servers to CSRS, I get a backtrace on the new CSRS second and third config servers. This happens about 4 seconds after reenabling the balancer. I am using MongoDB version 3.2.1-95-g4a3c6e6 Attached are the logs for the 3 config servers. run9007 is the first config server, run9008 the second, and run9009 the third. The balancer was restored around the time 11:19:01. The backtraces for run9008 and run9009 occur around the time 11:19:05. |
| Comments |
| Comment by Githook User [ 29/Jan/16 ] | |||||||||||||||||||||||||||||||||||
|
Author: {u'username': u'scotthernandez', u'name': u'Scott Hernandez', u'email': u'scotthernandez@gmail.com'}Message: | |||||||||||||||||||||||||||||||||||
| Comment by Githook User [ 29/Jan/16 ] | |||||||||||||||||||||||||||||||||||
|
Author: {u'username': u'scotthernandez', u'name': u'Scott Hernandez', u'email': u'scotthernandez@gmail.com'}Message: (cherry picked from commit 1175510702d070e4eeed436d81c52dec855f664e) | |||||||||||||||||||||||||||||||||||
| Comment by Scott Hernandez (Inactive) [ 29/Jan/16 ] | |||||||||||||||||||||||||||||||||||
|
I believe the failure is that the getmore returns with a cursorId of 0, indicating there is no more, and not sending a BSONObjectBuilder pointer, which the invariant hits. This was not a case the code was built to support so we need to test for this case and allow it – also, invariant is wrong here I believe either way since the code can recovery by issuing a new query/fetcher. | |||||||||||||||||||||||||||||||||||
| Comment by Scott Hernandez (Inactive) [ 28/Jan/16 ] | |||||||||||||||||||||||||||||||||||
|
Out until Monday, pass back if | |||||||||||||||||||||||||||||||||||
| Comment by Spencer Brody (Inactive) [ 28/Jan/16 ] | |||||||||||||||||||||||||||||||||||
|
Looks like a failure in bgsync, Scott can you take a look? | |||||||||||||||||||||||||||||||||||
| Comment by Spencer Brody (Inactive) [ 28/Jan/16 ] | |||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||
| Comment by Ramon Fernandez Marina [ 28/Jan/16 ] | |||||||||||||||||||||||||||||||||||
|
From the steps to reproduce:
|