Hi Jon, Thanks for pointing that out. Looks like I missed that during the CASSANDRA-9608 review. Robert and I didn't discuss that online or offline, so it's something that snuck in.
I don't have a personal preference for GC algs right now, but at a minimum it would probably be appropriate to return to CMS. We can certainly have a discussion thread about switching the default, but until we have that that, I think it's more honest to revert to CMS. If there's no immediate push back on reverting, I'll have a patch within a day to revert. Thanks, -Jason On Wed, Aug 8, 2018 at 3:34 PM, Jonathan Haddad <j...@jonhaddad.com> wrote: > I fired up trunk to check something, and noticed this: > > INFO [Service Thread] 2018-08-08 15:01:36,723 GCInspector.java:290 - G1 > Young Generation GC in 339ms. G1 Eden Space: 4634705920 -> 0; G1 Old Gen: > 1190138352 -> 1435504616; G1 Survivor Space: 406847488 -> 301989888; > > which I thought was a bit weird, since I was using trunk without making any > changes and didn't remember seeing a JIRA where we agreed to make that > change. I looked back and saw it made it in as a side effect of > CASSANDRA-9608, but wasn't explicitly discussed in the ticket, and there's > no note of it in CHANGES. > > I'm personally OK with this change as G1 is a safer bet for anyone who uses > the defaults, but we should be explicit about the change. Can anyone think > of a reason why we'd need to revert this back to ParNew / CMS? > > -- > Jon Haddad > http://www.rustyrazorblade.com > twitter: rustyrazorblade >
