log-explorer Helm values reference#
Helm values reference table for the log-explorer application.
Key |
Type |
Default |
Description |
|---|---|---|---|
affinity |
object |
|
Affinity rules for the log-explorer deployment pod |
cache |
object |
See |
Persistent volume for the log cache. Fetching a night out of Loki takes minutes, so the cache is what makes the tool usable a second time; an emptyDir would throw it away on every pod restart. |
cache.mountPath |
string |
|
Where the cache volume is mounted, and what the application is told to use as its cache root. |
cache.sizeGi |
int |
|
Size of the cache volume, in GiB. Given as a bare number rather than a quantity string because it sets two things that must not disagree: the volume’s size, and the byte ceiling the application’s LRU eviction keeps the cache under ( |
cache.storageClassName |
string |
|
Storage class for the cache volume |
cache.usagePercent |
int |
|
Percentage of the volume the application will fill before it starts evicting least-recently-used windows. The headroom covers filesystem overhead and the gap between a fetch finishing and the eviction sweep that follows it — the cache is briefly over its ceiling in between. |
config.lokiPassword |
string |
|
Loki password in the clear, for bringing an environment up before its Vault secret exists. Do not commit a real value: this repository is public, and anything set here also shows in plain text in |
config.lokiUsername |
string |
|
Loki HTTP basic-auth username. The password comes from the |
config.site |
object |
See the per-environment values files |
The one site this deployment serves. A site pairs a Loki cluster (where the logs live) with the ConsDB endpoint that owns shutter-close truth for that cluster’s data; the two must match or a dataId resolves against the wrong observatory. Singular on purpose: an instance on manke is BTS and one on yagan is the summit, and because the same 13-digit dataId exists at both with different obs_end values, an instance that could be pointed at the other one would return answers that looked plausible rather than obviously wrong. Needs |
config.windowAfterS |
int |
|
Seconds after shutter close that a fetch window ends. As above, a starting value rather than a cap. Worth raising where the pipeline runs slower than the exposure cadence. |
config.windowBeforeS |
int |
|
Seconds before shutter close that a fetch window starts. This is the starting value of an editable field, not a cap: widening a window to catch a neighbouring exposure is a normal investigative move. |
config.workers |
int |
|
Parallel log-download workers per fetch. Each is a |
global.host |
string |
Set by Argo CD |
Host name for ingress |
global.vaultSecretsPath |
string |
Set by Argo CD |
Base path for Vault secrets |
image.pullPolicy |
string |
|
Pull policy for the log-explorer image |
image.repository |
string |
|
Image to use in the log-explorer deployment |
image.tag |
string |
The appVersion of the chart |
Tag of image to use |
imagePullSecrets |
list |
See |
Image pull secrets. Needed only if the GHCR package is made private; a public package pulls without credentials. |
ingress.annotations |
object |
|
Additional annotations for the ingress rule |
ingress.path |
string |
|
Path the app is served under. Passed to the application as |
nodeSelector |
object |
|
Node selection rules for the log-explorer deployment pod |
podAnnotations |
object |
|
Annotations for the log-explorer deployment pod |
replicaCount |
int |
|
Number of web deployment pods to start. Must stay at 1: the log cache is a ReadWriteOnce volume, and loaded exposures live in the serving process’s memory, so a second replica would answer with different state depending on which pod the request landed on. |
resources |
object |
See |
Resource limits and requests for the log-explorer deployment pod |
tolerations |
list |
|
Tolerations for the log-explorer deployment pod |