fix: update ceph-cluster + loki-fluentbit + log for 2026-07-25 fixes (OSD purge, replication_factor, CNPG leader election)

This commit is contained in:
Dominik Schön
2026-07-25 09:27:19 +00:00
parent dc316f7b26
commit 92a13fdfa4
3 changed files with 26 additions and 8 deletions
+11 -2
View File
@@ -3,15 +3,24 @@ title: Loki + Fluent Bit (Logging Stack)
category: systems
tags: [loki, fluent-bit, logging, observability]
created: "2026-07-24"
modified: "2026-07-24"
modified: "2026-07-25"
---
# Loki + Fluent Bit (Logging Stack)
## Loki
- **Namespace:** logging (K8s)
- **Mode:** SingleBinary, `replication_factor: 1` (fix 2026-07-25)
- **Chart:** grafana/loki v6.42.0 via ArgoCD GitOps
- **Caches:** memcached disabled (`chunksCache.enabled: false`, `resultsCache.enabled: false`)
- **External Service:** `loki-external`, LoadBalancer `10.0.30.207:3100`
- **Caches:** memcached (results-cache, `-m 512 -I 2m`, requests 256Mi/limits 512Mi)
### Known Issues (Fixed 2026-07-25)
| Issue | Root Cause | Fix |
|-------|-----------|-----|
| HTTP 500 on push | `replication_factor: 3` in hash ring with only 1 instance → "too many unhealthy instances in the ring" | `loki.commonConfig.replication_factor: 1` (NOT `loki.common` — chart ignores that path) |
| Memcached CrashLoop | Caches allocated 0 memory | Disabled caches entirely (SingleBinary doesn't need them) |
| Fluent Bit connection failures | Cascading from Loki 500s | Fixed by Loki replication_factor fix |
## Fluent Bit v5.0.9