Skip to content

Resolve "instrument signald read path for APM metrics"

aguestuser requested to merge 324-instrument-signald-read-path-for-apm-metrics into master

Closes #324 (closed)

context:

  • we're starting to get decent metrics for when and why sendMessage takes longer
  • in observing those metrics we've noticed that message sending latency shoots up when there are a bunch of incoming read receipts. we'd like to better understand why. this is a first pass at instrumenting the signald read path to see what takes a long time when processing incoming messages

changes:

  • match team-friendo signald fork mr-13: signald-fork!13 (merged)
  • ie: "add a bunch of histogram timers to Manager.receiveMessages to try to understand which parts of that method consume the most time/resources"
  • ie: pin our signald dockerfile to:

side-effects

pull latest signalboost/signald images at top of deploy

  • the first invocation of docker-compose would have done this anyway but by doint it as its own step, we compartmentalize the pull into its own step, thereby avoid confusingly long response times for steps (like installing node packages) that we expect to take a short time when they have little work to do

Merge request reports