MongoDB - Defect ID: 420939
ReplSetTest.remove() should call stop() on the node to be removed

Info

d20002| 2017-08-28T10:35:36.613-0400 E STORAGE [thread2] WiredTiger error (2) [1503930936:613065][16730:0x7f3029f2c700], log-server: /data/db/__unknown_name__-2/journal: directory-list: opendir: No such file or directory d20002| 2017-08-28T10:35:36.613-0400 E STORAGE [thread2] WiredTiger error (2) [1503930936:613149][16730:0x7f3029f2c700], log-server: log pre-alloc server error: No such file or directory d20002| 2017-08-28T10:35:36.613-0400 E STORAGE [thread2] WiredTiger error (2) [1503930936:613161][16730:0x7f3029f2c700], log-server: log server error: No such file or directory d20002| 2017-08-28T10:35:36.613-0400 E STORAGE [thread2] WiredTiger error (-31804) [1503930936:613171][16730:0x7f3029f2c700], log-server: the process must exit and restart: WT_PANIC: WiredTiger library panic d20002| 2017-08-28T10:35:36.613-0400 F - [thread2] Fatal Assertion 28558 at src/mongo/db/storage/wiredtiger/wiredtiger_util.cpp 361 d20002| 2017-08-28T10:35:36.613-0400 F - [thread2] d20002| d20002| ***aborting after fassert() failure This crash occurs after the following steps: $ ./mongo --nodb MongoDB shell version v0.0.0 > var replTest = new ReplSetTest({nodes: 1}); > replTest.startSet(); // start mongod > replTest.remove(0); // Remove it from replsettest's config > replTest.stopSet(); // crashes mongod.

Top User Comments

xgen-internal-githook commented on Tue, 5 May 2020 23:20:52 +0000: Author: {'name': 'Jason Chan', 'email': 'jason.chan@10gen.com', 'username': 'jasonjhchan'} Message: SERVER-30862 ReplSetTest.remove() should call stop() on the node to be removed Branch: master https://github.com/mongodb/mongo/commit/ec12de9fcd39f4f7eff7c98dde4b3de6ff806778 jason.chan commented on Thu, 30 Apr 2020 18:20:51 +0000: I was able to reproduce this on v4.5. I believe it makes sense to have remove() call stop() on the node. siyuan.zhou@10gen.com commented on Thu, 9 Apr 2020 06:10:24 +0000: In Safe Reconfig project, we usually restore the config to the original config. I think one reason is to avoid problems like this one. We should make ReplSetTest smarter to handle removed nodes. jack.mulrow commented on Fri, 15 Sep 2017 16:35:51 +0000: I accidentally committed SERVER-30682 with this ticket's number, then I reverted it when I realized my mistake, so please ignore the above two comments. ramon.fernandez commented on Fri, 15 Sep 2017 16:11:44 +0000: Author: {'username': u'jsmulrow', 'name': u'Jack Mulrow', 'email': u'jack.mulrow@mongodb.com'} Message:SERVER-30862 Run the concurrency suite with "secondary" read preference Branch:master https://github.com/mongodb/mongo/commit/6dded939c14a072d6b15a47a692ba13c706d8db1 ramon.fernandez commented on Fri, 15 Sep 2017 16:11:42 +0000: Author: {'username': u'jsmulrow', 'name': u'Jack Mulrow', 'email': u'jack.mulrow@mongodb.com'} Message:Revert "SERVER-30862 Run the concurrency suite with "secondary" read preference" This reverts commit 6dded939c14a072d6b15a47a692ba13c706d8db1. Branch:master https://github.com/mongodb/mongo/commit/cbfcf83f39a05a6921d68a0b7350e273f86cbd22 milkie commented on Mon, 28 Aug 2017 15:42:00 +0000: It turns out that ReplSetTest.remove(foo) cannot be called before first calling ReplSetTest.stop(foo). tags2.js is one such test that is following the rules. If you don't stop a node prior to removing it, when stopSet runs it forgets to stop the removed node, and then proceeds to delete the dbpath out from under it. We should either remove the remove() function from ReplSetTest, or add a call to stop() as the first thing that function does.

Steps to Reproduce