Graph Odoo Request Duration with Loki#

Before trying to improve performance, it is a good idea to measure. We want a graph that shows us request durations segmented by query time (the time it takes Postgres to gather the data) and remaining time (how long it takes Python/Odoo to do things with the data), like this:

graph

Understanding the Log Lines#

Odoo 15 defines the log format as

%(asctime)s %(pid)s %(levelname)s %(dbname)s %(name)s: %(message)s %(perf_info)s

with perf_info defined as

query_count, query_time, remaining_time

So a log line like this

2023-10-17 10:20:45,053 1707305 INFO odoo werkzeug: 1.2.3.4 - - [17/Oct/2023 10:20:45] "POST /web/dataset/search_read HTTP/1.0" 200 - 20 0.012 0.033

Gives you:

field

value

asctime

2023-10-17 10:20:45,053

pid

1707305

levelname

INFO

dbname

odoo

name

werkzeug

message

1.2.3.4 - - [17/Oct/2023 10:20:45] "POST /web/dataset/search_read HTTP/1.0" 200 -

query_count

20

query_time

0.012

remaining_time

0.033

Note:

Loki Pattern Query#

Use Lokis pattern parser to parse lines into fields:

{job="varlogs",filename="/var/log/daemon.log"}
  | pattern `<dt> <time> <pid> <level> <dbname> <logger>: <ip> <_> <_> [<_>] "<method> <path> <_>" <status> <_> <queries> <query_seconds> <remaining_seconds>`

Graphing#

Our first Series unwraps by query_seconds and aggregates via max_over_time over $__interval. The outside max function aggregates over the remaining dimensions, as we are only interested in the highest value of the duration in query_seconds.

max(max_over_time(
  {job="varlogs",filename="/var/log/daemon.log"}
  | pattern `<dt> <time> <pid> <level> <dbname> <logger>: <ip> <_> <_> [<_>] "<method> <path> <_>" <status> <_> <queries> <query_seconds> <remaining_seconds>`
  | unwrap query_seconds
  [$__interval]
))

Then

  • add another query for remaining_seconds

  • set title to request durations max [$__interval]

  • set bar chart

  • set stack series: normal

  • set unit to seconds (s)

An that’s it! 📊

[unwrap]:

Odoo Format Fields   Hetzner Cloud Volume for Docker Volume Storage