Dinesh Bhat has posted comments on this change.

Change subject: KUDU-1741: Keep MiniCluster::Restart consistent with 
ExternalMiniCluster::Restart
......................................................................


Patch Set 2:

> > Personally I think this patch is fine as-is, because the
 > particular
 > > semantics matter a bit less than just being consistent IMHO. If
 > we
 > > wanted to change the semantics to work the other way around, we
 > > could do it, but it would be more work to get right since EMC
 > gets
 > > way more use than MC.
 > 
 > Agreed with Mike; consistency takes precedence here.
 > 
 > I took a quick look at the various users of Restart() today (both
 > EMC and the individual daemons), and it's important to note that
 > quite a few tests do something like this:
 > 
 > <call Shutdown() to shut something down>
 > <do some work>
 > <call Restart() to bring it back up>
 > 
 > Thus, if we took this in the other direction and made Restart()
 > also shut things down, we wouldn't be able to handle the above use
 > case. As another data point, linked_list-test uses ShutdownNodes()
 > to bring down just the tservers, then Restart() to bring them back
 > up.
 > 
 > Personally, I find the proliferation of Start/Stop methods to be
 > confusing when writing new tests, so my vote would be for 1)
 > convert Restart() into stop-then-start semantics as JD suggested,
 > 2) beef up Start() such that it's idempotent and will bring up all
 > downed servers, and 3) ensure these semantics are consistent not
 > just across EMC/MC but also on the individual daemons themselves
 > (i.e. Start() should be idempotent everywhere).
 > 
 > With the above changes, the back-to-back cluster Shutdown()+Restart()
 > callers become single calls to Restart(), and the more interesting
 > cases become <shutdown some stuff>+<do some work>+Start().
 > 
 > But, I'm also fine with deferring this (more extensive) change to
 > later, and to focus on bringing in consistency now.

Thanks Adar, I will track these discussions/ideas in a new JIRA for future 
improvements.

-- 
To view, visit http://gerrit.cloudera.org:8080/5598
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: Iad33b7c46bfca3f277ccbca7d0420272f06a6633
Gerrit-PatchSet: 2
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Dinesh Bhat <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Dinesh Bhat <[email protected]>
Gerrit-Reviewer: Jean-Daniel Cryans <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Mike Percy <[email protected]>
Gerrit-HasComments: No

Reply via email to