Is it normal for sql.txn.abort.count to reach several thousand/hour?

deployment
#1

I noticed that the sql.txn.abort.count “seems” high on my cluster, averaging around 5000 per hour. It looks to be independent of load, even with no queries at all it’s at roughly that level. Looking back there has not been any major variance in this since I started the cluster.

This is on a cluster with 3 nodes on kubernetes on Azure.

Should this be cause for concern or is it normal?

0 Likes

(Ron Arévalo) #2

Hey @sebastianblunt,

What version of CRDB are you running? Are you able to provide us with your DDL and DML?

Also, if you go to the statements detail page could you see how many failed queries are being reported?

For reference:

Thanks,

Ron

0 Likes

#3

Hi Ron,

So I was reading this off of datadog, which for some reason shows the count as going up and down. When I read the metric directly from the cockroach admin UI it’s a constant (on a short timeframe) which confused me, but then I realized it was showing count and not rate. The rate is correctly at a constant zero.

For some reason datadog is showing crazy ups and downs while the cockroach admin UI is showing no queries, so I think this is a bug with datadog and not with cockroachdb.

Thanks for responding! I’ll be sure to write if I have any further questions :slight_smile:

0 Likes

(Ron Arévalo) #4

Hey @sebastianblunt,

Well that’s good to know!

Have a good weekend!

Ron

0 Likes