Currently two data providers collect statistics on how often a remote cache was checked, and how many were hits/misses. One of the two can be dropped. CachingAndExecutionMetrics (new) seems to do a better job at it than RemoteCacheMetrics.
Example profile collected from a fully incremental build using --disk_cache:
CachingAndExecutionMetrics: Metrics on the processed actions' caching and execution status.
Actions: ____________________________ 3869
Remote cache hits: ______________ 3541 91,52% of all actions
Remote cache misses: ____________ 0 0,00% of all actions
Executed locally: 0 0,00% of all cache misses
Executed remotely: 0 0,00% of all cache misses
Not reported: 0 0,00% of all cache misses
Remote cache not checked: _______ 302 7,81% of all actions
Executed locally: 3 0,99% of all cache skips
Executed remotely: 0 0,00% of all cache skips
Internal (not exhaustive): 26 8,61% of all cache skips
Not reported 299 99,01% of all cache skips
Executed locally: _______________ 3 0,08% of all actions
Executed remotely: ______________ 0 0,00% of all actions
Execution not reported: _________ 3866 99,92% of all actions
RemoteCacheMetrics: Metrics on the remote caching used. This includes both the use of `--remote_cache` and `--disk_cache`.
Number of cache checks: 3540
Number of cache misses: 2313
Cache hit percentage: 34,66%
Time spend checking for cache hits: 11s
Time spend downloading outputs: 209ms
Time spend uploading outputs: 0ms
CachingAndExecutionMetrics matches with the Bazel output on disk cache hit, so we should remove the cache check evaluation from RemoteCacheMetrics and keep the data collected by CachingAndExecutionMetrics
Currently two data providers collect statistics on how often a remote cache was checked, and how many were hits/misses. One of the two can be dropped.
CachingAndExecutionMetrics
(new) seems to do a better job at it thanRemoteCacheMetrics
.Example profile collected from a fully incremental build using
--disk_cache
:Bazel output:
INFO: 3869 processes: 3541 disk cache hit, 325 internal, 2 darwin-sandbox, 1 local.
BIA:
CachingAndExecutionMetrics
matches with the Bazel output ondisk cache hit
, so we should remove the cache check evaluation fromRemoteCacheMetrics
and keep the data collected byCachingAndExecutionMetrics