diff options
| author | Paul Buetow <paul@buetow.org> | 2026-07-26 00:35:39 +0300 |
|---|---|---|
| committer | Paul Buetow <paul@buetow.org> | 2026-07-26 00:35:39 +0300 |
| commit | 424a3cf07654a9a323cf8591c2ebf9ee177a3fdd (patch) | |
| tree | 2167197b84b0cd59a0b3a35185d57227e6b6b43d /f3s/git-server/helm-chart/templates/deployment.yaml | |
| parent | de406f3145b07a6bfdd62d2e223075d9d20fa1eb (diff) | |
git-server: cut cgit render time by enabling its response cache
cgit ran with cache-size=0 against scan-path over ~80 repos on NFS, so every
request reopened all of them to resolve the index age column. Measured 376ms
per index render, of which only ~50ms was CPU -- the rest was NFS wait. With
the cache enabled the same render takes ~19-29ms.
cache-root points at the existing node-local cgit-runtime emptyDir rather than
the NFS-backed /repos; caching onto NFS is what the old "permission issues"
comment was actually about. The startup script creates the directory since
cgit does not.
fcgiwrap ran as a single worker, so the 26 ArgoCD apps polling conf.git through
git-http-backend serialized against web UI renders; 4 concurrent requests
managed only ~1.5x the throughput of one. Preforking 4 workers separates them.
The cgit container also idled at 105m against a 250m limit and had accumulated
~2.9h of throttled time (nr_throttled 96979), so raise the ceiling and set the
request from measured idle usage.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Diffstat (limited to 'f3s/git-server/helm-chart/templates/deployment.yaml')
| -rw-r--r-- | f3s/git-server/helm-chart/templates/deployment.yaml | 26 |
1 files changed, 21 insertions, 5 deletions
diff --git a/f3s/git-server/helm-chart/templates/deployment.yaml b/f3s/git-server/helm-chart/templates/deployment.yaml index 28733c5..4206cd8 100644 --- a/f3s/git-server/helm-chart/templates/deployment.yaml +++ b/f3s/git-server/helm-chart/templates/deployment.yaml @@ -204,8 +204,19 @@ spec: }\ ' /tmp/conf.d/default.conf - # Start fcgiwrap with socket in /tmp - spawn-fcgi -s /tmp/fcgiwrap.sock -n -- /usr/bin/fcgiwrap & + # cgit's response cache lives here (see cache-root in the cgit-config + # ConfigMap). It must exist before cgit runs -- cgit does not create it. + # /tmp is the cgit-runtime emptyDir: node-local, so cache writes never + # go back over NFS. The cache is intentionally per-pod and cold after + # a restart. + mkdir -p /tmp/cgit-cache + + # Start fcgiwrap with socket in /tmp. + # -c 4 preforks 4 workers. With a single worker the 26 ArgoCD apps + # polling conf.git through git-http-backend serialize against web UI + # renders on the same CGI process; measured 4 concurrent requests at + # only ~1.5x the throughput of one. + spawn-fcgi -s /tmp/fcgiwrap.sock -n -- /usr/bin/fcgiwrap -c 4 & sleep 1 chmod 666 /tmp/fcgiwrap.sock exec nginx -c /tmp/nginx.conf -g 'daemon off;' @@ -239,13 +250,18 @@ spec: allowPrivilegeEscalation: false capabilities: drop: ["ALL"] + # cgit idled at 105m against the previous 250m limit -- 42% of its own + # ceiling at rest -- and the cgroup had accumulated ~2.9h of throttled + # time (nr_throttled 96979). Renders are bursty, so the quota was being + # hit on ordinary page loads. Headroom raised for the 4 fcgiwrap + # workers; the request tracks measured idle usage. resources: requests: - cpu: 50m + cpu: 100m memory: 128Mi limits: - cpu: 250m - memory: 256Mi + cpu: 1000m + memory: 512Mi volumes: - name: repos |