https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
Gabriel Wicke gwi...@wikimedia.org changed:
What|Removed |Added
Status|ASSIGNED|RESOLVED
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #7 from Marc Ordinas i Llopis marc...@wikimedia.org ---
Just a note on why the code seems different to
https://github.com/felixge/node-mysql#transactions : The transactions support
in node-mysql is very recent and (afaik) hasn't
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #8 from Marc Ordinas i Llopis marc...@wikimedia.org ---
After following retained memory in gwicke's snapshot, it seems the biggest
culprit is the _callSite member of Query. Lo and behold, then I found this
commit for node-mysql:
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #9 from Gabriel Wicke gwi...@wikimedia.org ---
It was at 2.0.0-alpha8. It looks like upgrading to 2.0.0-alpha9 has helped.
--
You are receiving this mail because:
You are on the CC list for the bug.
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #10 from Gabriel Wicke gwi...@wikimedia.org ---
Looks better, but:
[14:24] gwicke mysql on the rt server does not look too happy
[14:24] gwicke the node portion of it is now at 380M
[14:24] gwicke probably a lot of waiting /
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #5 from Marc Ordinas i Llopis marc...@wikimedia.org ---
I did detect some amount of memory leakage while profiling locally, if I recall
correctly, ~90k per fetch/results cycle. As I mentioned in a previous comment,
the leakage
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #6 from Gabriel Wicke gwi...@wikimedia.org ---
I took a heap snapshot in production and inspected it in chromium. It seems
that a large number of closures are responsible for the memory leak. From
browsing the data a bit it seems
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #3 from ssas...@wikimedia.org ---
Gabriel: are you sure this is the rt server? Here is latest top from which I
extracted nodejs/node info. This seems like the parsoid web service clogged up.
The RT server command is 'node' and
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #4 from ssas...@wikimedia.org ---
I see it now as well ... 1.5G so far. t So, there are probably different
sources -- some for workers, some for the RT server. Hmm ...
I got confused because when I looked at top at that time, the
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
--- Comment #2 from Gabriel Wicke gwi...@wikimedia.org ---
Just happened again. Had to kill -9 it, normal restart did not help. top showed
1.4G rss.
--
You are receiving this mail because:
You are on the CC list for the bug.
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
Marc Ordinas i Llopis marc...@wikimedia.org changed:
What|Removed |Added
Status|NEW |ASSIGNED
https://bugzilla.wikimedia.org/show_bug.cgi?id=53776
Gabriel Wicke gwi...@wikimedia.org changed:
What|Removed |Added
Priority|Unprioritized |High
--
You are
12 matches
Mail list logo