mirror of
https://github.com/coder/coder.git
synced 2026-06-03 04:58:23 +00:00
Callum Styan
45c43d4ec4
This should resolve https://github.com/coder/internal/issues/728 by refactoring the ResourceMonitorAPI struct to only require querying the resource monitor once for memory and once for volumes, then using the stored monitors on the API struct from that point on. This should eliminate the vast majority of calls to `GetWorkspaceByAgentID` and `FetchVolumesResourceMonitorsUpdatedAfter`/`FetchMemoryResourceMonitorsUpdatedAfter` (millions of calls per week). Tests passed, and I ran an instance of coder via a workspace with a template that added resource monitoring every 10s. Note that this is the default docker container, so there are other sources of `GetWorkspaceByAgentID` db queries. Note that this workspace was running for ~15 minutes at the time I gathered this data. Over 30s for the `ResourceMonitor` calls: ``` coder@callum-coder-2:~/coder$ curl localhost:19090/metrics | grep ResourceMonitor | grep count % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 0coderd_db_query_latencies_seconds_count{query="FetchMemoryResourceMonitorsByAgentID"} 2 coderd_db_query_latencies_seconds_count{query="FetchMemoryResourceMonitorsUpdatedAfter"} 2 100 288k 0 288k 0 0 58.3M 0 --:--:-- --:--:-- --:--:-- 70.4M coderd_db_query_latencies_seconds_count{query="FetchVolumesResourceMonitorsByAgentID"} 2 coderd_db_query_latencies_seconds_count{query="FetchVolumesResourceMonitorsUpdatedAfter"} 2 coderd_db_query_latencies_seconds_count{query="UpdateMemoryResourceMonitor"} 155 coderd_db_query_latencies_seconds_count{query="UpdateVolumeResourceMonitor"} 155 coder@callum-coder-2:~/coder$ curl localhost:19090/metrics | grep ResourceMonitor | grep count % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 0coderd_db_query_latencies_seconds_count{query="FetchMemoryResourceMonitorsByAgentID"} 2 coderd_db_query_latencies_seconds_count{query="FetchMemoryResourceMonitorsUpdatedAfter"} 2 100 288k 0 288k 0 0 34.7M 0 --:--:-- --:--:-- --:--:-- 40.2M coderd_db_query_latencies_seconds_count{query="FetchVolumesResourceMonitorsByAgentID"} 2 coderd_db_query_latencies_seconds_count{query="FetchVolumesResourceMonitorsUpdatedAfter"} 2 coderd_db_query_latencies_seconds_count{query="UpdateMemoryResourceMonitor"} 158 coderd_db_query_latencies_seconds_count{query="UpdateVolumeResourceMonitor"} 158 ``` And over 1m for the `GetWorkspaceAgentByID` calls, the majority are from the workspace metadata stats updates: ``` coder@callum-coder-2:~/coder$ curl localhost:19090/metrics | grep GetWorkspaceByAgentID | grep count % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 284k 0 284k 0 0 42.4M 0 --:--:-- --:--:-- --:--:-- 46.3M coderd_db_query_latencies_seconds_count{query="GetWorkspaceByAgentID"} 876 coder@callum-coder-2:~/coder$ curl localhost:19090/metrics | grep GetWorkspaceByAgentID | grep count % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 284k 0 284k 0 0 75.4M 0 --:--:-- --:--:-- --:--:-- 92.7M coderd_db_query_latencies_seconds_count{query="GetWorkspaceByAgentID"} 918 ``` --------- Signed-off-by: Callum Styan <callumstyan@gmail.com>