[g3-canary] Add PROD.md for infra failure alerts
Bug: skia:12521
Change-Id: I4620aafdf1a1d83039e4a23ede53e2a3d2351d55
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/457760
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>
diff --git a/infra/bots/task_drivers/g3_canary/PROD.md b/infra/bots/task_drivers/g3_canary/PROD.md
new file mode 100644
index 0000000..b0464d2
--- /dev/null
+++ b/infra/bots/task_drivers/g3_canary/PROD.md
@@ -0,0 +1,20 @@
+G3 Canary Production Manual
+===========================
+
+General information about canaries is available in
+[go/autoroller-canary-bots](https://goto.google.com/autoroller-canary-bots).
+
+Alerts
+======
+
+g3_canary_infra_failures
+------------------------
+
+Happens when the skia_try_service in G3 returns an exception.
+Check the skia_try_service's error logs in [go/skia-borg-jobs](go/skia-borg-jobs).
+
+For errors that do not seem to be transient, restarting the borg job has worked
+in the past:
+```
+borg --borg=${BORG_CELL} --user=skia --name=skia_try_service --avoid_parent restarttask 0
+```
