From: Vladislav Shpilevoy <v.shpilevoy@tarantool.org> To: Ilya Kosarev <i.kosarev@tarantool.org>, tarantool-patches@dev.tarantool.org Subject: Re: [Tarantool-patches] [PATCH v2 2/4] replication: fix appliers pruning Date: Sun, 24 Nov 2019 16:54:03 +0100 [thread overview] Message-ID: <8b2b2757-8926-1fbc-3f2b-1a1afea6fa52@tarantool.org> (raw) In-Reply-To: <66ff4c73f329f83b616eeadbf5ea6f26994ecb4a.1574545667.git.i.kosarev@tarantool.org> Thanks for the patch! On 23/11/2019 22:53, Ilya Kosarev wrote: > During pruning of replicaset appliers some anon replicas might connect > from replicaset_follow called in another fiber. Therefore we need to > prune appliers of anon replicas first. > > Closes #4643 > --- > src/box/replication.cc | 14 +++++++------- > 1 file changed, 7 insertions(+), 7 deletions(-) > > diff --git a/src/box/replication.cc b/src/box/replication.cc > index 509187cd3..3ccfa8b33 100644 > --- a/src/box/replication.cc > +++ b/src/box/replication.cc > @@ -516,23 +516,23 @@ replicaset_update(struct applier **appliers, int count) > */ > > /* Prune old appliers */ > - replicaset_foreach(replica) { > - if (replica->applier == NULL) > - continue; > + rlist_foreach_entry_safe(replica, &replicaset.anon, in_anon, next) { 'safe' rlist iteration won't help, if a next anon replica will connect during yield in applier_stop(). Safe iteration can cope with deletion only of the current element. You can try while (rlist_empty(...)) { replica = rlist_first(...); ... } The same about the cycle below. And I am not really sure, that you need two cycles now. Just keep it one, but iterate like I showed above. > applier = replica->applier; > replica_clear_applier(replica); > - replica->applier_sync_state = APPLIER_DISCONNECTED; > + replica_delete(replica); > applier_stop(applier); > applier_delete(applier); > } > - rlist_foreach_entry_safe(replica, &replicaset.anon, in_anon, next) { > + rlist_create(&replicaset.anon); > + replicaset_foreach(replica) { > + if (replica->applier == NULL) > + continue; > applier = replica->applier; > replica_clear_applier(replica); > - replica_delete(replica); > + replica->applier_sync_state = APPLIER_DISCONNECTED; > applier_stop(applier); > applier_delete(applier); > } > - rlist_create(&replicaset.anon); > > /* Save new appliers */ > replicaset.applier.total = count; >
next prev parent reply other threads:[~2019-11-24 15:54 UTC|newest] Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top 2019-11-23 21:53 [Tarantool-patches] [PATCH v2 0/4] fix box_set_replication issues & stabilize quorum test Ilya Kosarev 2019-11-23 21:53 ` [Tarantool-patches] [PATCH v2 1/4] test: update test-run Ilya Kosarev 2019-11-23 21:53 ` [Tarantool-patches] [PATCH v2 2/4] replication: fix appliers pruning Ilya Kosarev 2019-11-24 15:54 ` Vladislav Shpilevoy [this message] 2019-11-23 21:53 ` [Tarantool-patches] [PATCH v2 3/4] replication: make anon replicas iteration safe Ilya Kosarev 2019-11-23 21:53 ` [Tarantool-patches] [PATCH v2 4/4] test: stabilize quorum test conditions Ilya Kosarev 2019-11-24 15:54 ` Vladislav Shpilevoy 2019-11-24 15:53 ` [Tarantool-patches] [PATCH v2 0/4] fix box_set_replication issues & stabilize quorum test Vladislav Shpilevoy 2019-11-25 11:50 ` Ilya Kosarev
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=8b2b2757-8926-1fbc-3f2b-1a1afea6fa52@tarantool.org \ --to=v.shpilevoy@tarantool.org \ --cc=i.kosarev@tarantool.org \ --cc=tarantool-patches@dev.tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH v2 2/4] replication: fix appliers pruning' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox