couchdb_httpd_up metric suggestion #92
equinsuocha
started this conversation in
General
Replies: 1 comment
-
Hey, thanks for pointing on that issue I was unaware of. I haven't had thorough thoughts about that behavior, so this is only a rough direction I'd like to go: the _up metric should express availability and also help understanding if the metrics in the response are complete and to be trusted. Yet, I do understand that "up" usually is more like a health check and should be simple and reliable. If you'd like to contribute a change, I'd prefer something like that:
|
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi there, I have some trouble understanding
couchdb_httpd_up
metric, so description tells us thatWas the last query of CouchDB stats successful.
, but if we look at exporter's code it's quite a complicated process with many possible points of failure, and practically when node stats collection fails for any reason (for example, in my case this error tends to happen from time to time: #30), I can see a strange situation whencouchdb_httpd_up == 0
, butcouchdb_httpd_node_up == 1
for everynode_name
, which makes wrong impression that service is down, however, actually it reflects some internal processing error which does not even affect node statuses and interrupt any other metric collection, and this metric is only showing that there was some error along the way and does not seem that important;So my question is, mb it makes sense to either change it to some general api health check, i.e. get "http://couch:5984" and just check response code, so we know that API is reachable and then if necessary introduce some internal metrics helping to understand that smth went wrong during
stats
collection, for example internal error counter labeled with error codes.If some of that makes sense and we can agree conceptually, I would like to contribute, since we are actively using your exporter in our enviromnent.
Beta Was this translation helpful? Give feedback.
All reactions