> On Aug. 20, 2020, 2:17 a.m., Qian Zhang wrote: > > src/slave/csi_server.cpp > > Lines 280 (patched) > > <https://reviews.apache.org/r/72779/diff/3/?file=2238503#file2238503line284> > > > > So here we erase the plugin while `plugin.initialized` is already > > associated at line 249 and `CSIServerProcess::publishVolume()` or > > `CSIServerProcess::unpublishVolume` may already be waiting on > > `plugins.at(name).initialized.future()`, will the erasing cause any > > problems?
We should be OK here since this callback is deferred onto the CSI server actor. If this `repair()` callback is invoked, that means that `plugins.at(name).initialized` has failed, which means that any callbacks chained onto `plugins.at(name).initialized` by `publishVolume()` or `unpublishVolume()` will be executed. By the time the deferred `repair()` callback is executed on the CSI server actor, failures will have already been returned for any previously-pending publish/unpublish calls. Does that make sense? - Greg ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/72779/#review221653 ----------------------------------------------------------- On Aug. 20, 2020, 11:14 p.m., Greg Mann wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/72779/ > ----------------------------------------------------------- > > (Updated Aug. 20, 2020, 11:14 p.m.) > > > Review request for mesos and Qian Zhang. > > > Repository: mesos > > > Description > ------- > > Initialized plugins lazily in the CSI server. > > > Diffs > ----- > > src/slave/csi_server.hpp f5ec7663926731483589c5e30e7060751ed01e55 > src/slave/csi_server.cpp 2ba4f22b92370b019722d845d6113fb37f1b876a > > > Diff: https://reviews.apache.org/r/72779/diff/4/ > > > Testing > ------- > > `sudo bin/mesos-tests.sh --gtest_filter="*CSIServerTest*"` > > > Thanks, > > Greg Mann > >
