promtorture: done

Author: ringerc

testcases/promtorture/.dockerignore

@@ -11,3 +11,4 @@ notes.md
 promtorture-tmp-*
 promtorture-metrics-*
 .gitignore
+metrics/

testcases/promtorture/.gitignore

@@ -4,3 +4,4 @@ promtool
 notes.md
 promtorture-metrics-*
 promtorture-tmp-*
+metrics/

testcases/promtorture/README.md

@@ -102,3 +102,14 @@ scripts/promcmd promtool query instant http://localhost:9090 'up{job="monitoring
 ```
 
 Other tips in `scripts/promcmd` comments.
+
+# Notes
+
+The apparent inability to force Prometheus to eagerly checkpoint its HEAD or to
+force a compaction is a bit of a pain; we can't easily analyse behaviour with
+querying data back from TSDB without waiting a couple of hours. We can force a
+snapshot, but that doesn't actually flush the HEAD.
+
+# See also
+
+* [prometheus docs on storage](https://github.com/juliusv/prometheus-docs/blob/master/content/docs/operating/storage.md)

testcases/promtorture/scripts/grab-metrics.sh

@@ -12,8 +12,8 @@ set -e -u -o pipefail -x
 
 source scripts/config
 
-tmpdir="promtorture-metrics-$(date -Isec)"
-mkdir "$tmpdir"
+tmpdir="metrics/promtorture-metrics-$(date -Isec)"
+mkdir -p "$tmpdir"
 
 echo 1>&2 "Dumping metrics to $tmpdir"