Merge ~woutervb/charm-graylog:master into ~graylog-charmers/charm-graylog:master

[Wouter van Bommel]

Make an option to specify the lag-time

[🤖 Canonical IS Merge Bot (canonical-is-mergebot)]

This merge proposal is being monitored by mergebot. Change the status to Approved to merge.

[Tom Haddon (mthaddon)]

How would a charm operator use this option and/or be aware of it?

review: Needs Information

9db4255... by Wouter van Bommel on 2019-05-20

Implemented the option indexer_failure_lag_hours

Via the charm config option: indexer_failure_lag_hours it possible to
specify how long back the python check should go when determining
indexer failures.
Defaults to 1 hour

[Junien F (axino)]

Just wanted to say that I'm +1 on the change from 1 day to 1 hour.

[Tom Haddon (mthaddon)]

Looks good, but please fix the lint error before landing.

review: Approve

75e5778... by Wouter van Bommel on 2019-05-22

Fix linting

[Junien F (axino)]

For the record, I don't agree that this should be a config option.

review: Disapprove

7907ba4... by Wouter van Bommel on 2019-06-10

Opinionate the charm

With this commit we force to look back for only 1 hour, when it comes to
index failures, and not the 24 hours like it was before

[Joe Guo (guoqiao)]

The current change is pretty simple now, looks good to me.

review: Approve

Unmerged commits

7907ba4... by Wouter van Bommel on 2019-06-10

Opinionate the charm

With this commit we force to look back for only 1 hour, when it comes to
index failures, and not the 24 hours like it was before

75e5778... by Wouter van Bommel on 2019-05-22

Fix linting

9db4255... by Wouter van Bommel on 2019-05-20

Implemented the option indexer_failure_lag_hours

Via the charm config option: indexer_failure_lag_hours it possible to
specify how long back the python check should go when determining
indexer failures.
Defaults to 1 hour

49adcee... by Wouter van Bommel on 2019-05-17

Make an option to specify the lag-time

By making the default time we look back for the index_failures a
parameter, defaulting to 1 hour, fixed indexing errors will be resolved
quicker by the service check.

It is a new option called --lag-time, specifying the look-back period in
hours, defaulting 1.

This will fix LP:1829467

diff --git a/files/check_graylog_health.py b/files/check_graylog_health.py
index 7fde420..ba780af 100755
--- a/files/check_graylog_health.py
+++ b/files/check_graylog_health.py
@@ -25,6 +25,7 @@ STATE_UNKNOWN = 3
 
 OVERALL_STATE = STATE_OK
 
+INDEX_LAG_HOURS = 1     # The amount of time we look back for indexing errors
 
 def try_check(name, gapi_con, args, output):
     global OVERALL_STATE
@@ -86,7 +87,8 @@ def outstanding_notifications(gapi_con, args):
 
 
 def indexer_failures(gapi_con, args):
-    since = datetime.datetime.utcnow() - datetime.timedelta(days=1)
+    since = datetime.datetime.utcnow() - \
+        datetime.timedelta(hours=INDEX_LAG_HOURS)
     failures = gapi_con.indexer_failures_count(since.isoformat())['count']
     msg = 'Indexer failures: {}'.format(failures)
     if failures == 0: