Skip to content

fix(logging): set flush_mode: interval on vmware_loginsight buffer#157

Merged
Varashi merged 1 commit into
mainfrom
fix/logging-buffer-flush-mode
May 27, 2026
Merged

fix(logging): set flush_mode: interval on vmware_loginsight buffer#157
Varashi merged 1 commit into
mainfrom
fix/logging-buffer-flush-mode

Conversation

@Varashi
Copy link
Copy Markdown
Owner

@Varashi Varashi commented May 27, 2026

Hotfix #4 for the logging migration. Default fluentd flush_mode is lazy → uses timekey 10m → flush_interval ignored. Sets explicit interval mode so chunks flush every 5s instead of sitting on disk for ~11 min.

Default fluentd buffer flush_mode is "lazy", which uses timekey-based
chunking — the operator's vmware-loginsight wrapper sets
`<buffer tag,time>` with `timekey 10m`/`timekey_wait 1m`, so chunks
sit on disk up to ~11 min before flushing to CFAPI. Our
`flush_interval: 5s` setting was logged as ignored.

Setting flush_mode: interval forces interval-based flushing with
the 5s cadence we set. End-to-end latency drops from ~11 min to
~5-10s (fluent-bit flush 1s + fluentd flush 5s + CFAPI POST + index).

Also adds flush_thread_count: 2 for parallel chunk POSTs — single-
threaded flushing was likely the next bottleneck if any chunk
stalled on a slow CFAPI response.

Surfaced live: synthetic 2464-byte stdout line emitted from a fresh
pod sat in the fluentd file buffer for ~5 min without reaching
vcflogs (buffer file present, no flush attempts in fluentd log).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@Varashi Varashi merged commit 0de6229 into main May 27, 2026
@Varashi Varashi deleted the fix/logging-buffer-flush-mode branch May 27, 2026 07:22
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant