Ori.livneh has uploaded a new change for review. https://gerrit.wikimedia.org/r/83230
Change subject: NavigationTiming StatsD instance: flush every 5 mins ...................................................................... NavigationTiming StatsD instance: flush every 5 mins The current flush interval (1 minute) isn't enough to collect enough events from the mobile site, and as a result the graph is very jumpy. Possible solutions: 1) Increase sampling rate across the board. Not desirable because the sampling rate for desktop is already more than adequate. 2) Selectively sample more mobile traffic. Not desirable because you'd then have to compensate for it when looking at Navigation Timing data in aggregate. Right now it applies the same sampling factor to all traffic, so you don't *have* to decompose it into mobile/desktop if you just want to consider site performance in aggregate. 3) Increase flush interval. This patch goes for #3, changing the flush interval from 1 min to 5 mins. Also included is a change to navtiming.py to make it log wiki-specific data. Change-Id: I0f24d071e148fe4765b520ea75f99b5bdb8a6371 --- M files/graphite/modules/navtiming.py M manifests/misc/graphite.pp 2 files changed, 5 insertions(+), 2 deletions(-) git pull ssh://gerrit.wikimedia.org:29418/operations/puppet refs/changes/30/83230/1 diff --git a/files/graphite/modules/navtiming.py b/files/graphite/modules/navtiming.py index b39cd3c..5a2babf 100644 --- a/files/graphite/modules/navtiming.py +++ b/files/graphite/modules/navtiming.py @@ -37,6 +37,7 @@ for meta in iter(zsock.recv_json, ''): if meta['revision'] != schema_rev: continue + wiki = meta['wiki'] event = meta['event'] if not event.get('isAnon'): continue @@ -48,3 +49,5 @@ if value > 0 and value < 60000: stat = 'browser.%s.%s:%s|ms' % (metric, site, value) sock.sendto(stat.encode('utf-8'), addr) + stat = 'browser.%s.%s.%s:%s|ms' % (metric, wiki, site, value) + sock.sendto(stat.encode('utf-8'), addr) diff --git a/manifests/misc/graphite.pp b/manifests/misc/graphite.pp index 418ccf4..46abffe 100644 --- a/manifests/misc/graphite.pp +++ b/manifests/misc/graphite.pp @@ -147,10 +147,10 @@ graphite_port => 2003, settings => { backends => [ './backends/graphite', './backends/ganglia' ], - flushInterval => 60 * 1000, # 1 min. + flushInterval => 5 * 60 * 1000, # 5 min. # Management API on loopback interface only. mgmt_address => $statsd_host, - percentThreshold => [ 5, 95 ], + percentThreshold => [ 95 ], # Show frequency distribution of client-side latency times. # See <http://tinyurl.com/statsd-histograms>. histogram => [ -- To view, visit https://gerrit.wikimedia.org/r/83230 To unsubscribe, visit https://gerrit.wikimedia.org/r/settings Gerrit-MessageType: newchange Gerrit-Change-Id: I0f24d071e148fe4765b520ea75f99b5bdb8a6371 Gerrit-PatchSet: 1 Gerrit-Project: operations/puppet Gerrit-Branch: production Gerrit-Owner: Ori.livneh <o...@wikimedia.org> _______________________________________________ MediaWiki-commits mailing list MediaWiki-commits@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mediawiki-commits