OPER-2071 Port Tapjoy changes to latest New Relic gem #3
Conversation
Forward-ported commit in question is here |
ping @jlogsdon since you were involved in the discussion earlier today |
This seems like a lot of changes to review. Wondering if it'd be easier to take the current gem and try to replay our change on top of it? |
Actually, looks like a pretty clean merge |
Mostly, yeah. I did have to resolve a conflict during rebase, but it was On Wednesday, March 2, 2016, StabbyCutyou notifications@github.com wrote:
Brent Stephens |
lgtm! Thanks for doing this! |
Should we do a burn-in test on chore for an hour or so, to see if the But this looked good in general
|
You could use |
It's probably worth it to at least do the |
I was going to ping Lionel, Alvin, or Jayanth since I think they've been working on chore-tester, but they don't have access to this repo so I can't do that. I'll drop them an email to see if they have time to help test. |
Jayanth is going to do some testing for us today. |
Should I change the build string to something a little more indicative of custom-ness? e.g. |
@StabbyCutyou @obrie @alanbrent Ran some tests using
Also, this was running on an Ubuntu machine, it read through 10000 messages, and the the number of open files were consistently between 62 and 64 during the entire execution. |
GIMME THE THUMB |
Would it be better for me to bring upstream into master with a separate PR? That would leave this PR with a single commit. |
My general strategy in the past has been to:
|
(updated comment since I hit submit a little too quickly) |
We add a 4th level to semver, i think, for managing our own version of someone elses gem. So if this were 1.2.3, our version would be 1.2.3.1 |
Last time we just incremented the build number by 1, which is what I've done On Friday, March 4, 2016, StabbyCutyou notifications@github.com wrote:
Brent Stephens |
The problem is that if the existing gem is 1.2.3, and we bump our copy to 1.2.4, it now sort of conflicts with the "Real" gem if that one ever bumps to 1.2.4. I recommend we tack on our own 4th semver particle, since it's A: compatible with the way bundler works and B: has no actual downside to it. |
Of course you reply after I've just finished building a TJS slug! And yeah, Brent Stephens On Fri, Mar 4, 2016 at 9:33 AM, StabbyCutyou notifications@github.com
|
Okay, so ... should we open a pull request upstream to obviate the need for customizing this gem? If so, I'm sure I can't be the one answering their inevitable questions :) |
Yeah I think we should be opening a PR, though I suspect that the change we've made is a workaround rather than a root cause. It'll take a little time to gather all of the information and examples needed to have a productive conversation -- but it's something we should definitely do. |
I am pretty sure we've tried pushing this upstream before and it wasn't accepted. |
FYI, the only upstream PR I found by us through a quick search was newrelic#135, which was a previous-to-this changeset and was accepted (yay!). Oddly enough, this patch we're forward-porting was committed the same day as that PR was opened. I imagine the intent was to go ahead and PR this one later, and it dropped out of memory. So ... seems best to leave this ball in @obrie's court. |
That was my understanding (according to my admittedly terrible memory) :) I'm happy to take on the responsibility of figuring out how to get this upstream. |
Blackmail would be my first suggestion On Mon, Mar 7, 2016 at 2:58 PM, Aaron Pfeifer notifications@github.com
|
…lizing threads for each ready pipe. This also prevents "Too many open files" errors from occurring
52c1ce0
to
6b42816
Compare
FYI the activity on this PR was me updating to a recently released update, finding out it breaks TJS specs, and reverting. We're back to the previous state, and this PR is ready for thumbs. |
This is required work for OPER-2045, related PR here
@Tapjoy/eng-group-ops