Bad syntax killing cockroach node?

General
#1

Hi,

We ran this query against a 3 node cockroach cluster

select avg(median),avg(perc98),avg(max) from api where url like '%/rsrv_eag/1/0/devices/{ID}/AccountVerificationRequest%' group by 1;

The query killed the node it was running on :(.

The complete table definition is

root@sl73commapd003.visa.com:9080> show columns from api;
+-----------+-----------+-------+----------------+
|   Field   |   Type    | Null  |    Default     |
+-----------+-----------+-------+----------------+
| id        | INT       | false | unique_rowid() |
| date      | TIMESTAMP | true  | NULL           |
| caller    | STRING    | true  | NULL           |
| component | STRING    | true  | NULL           |
| method    | STRING    | true  | NULL           |
| url       | STRING    | true  | NULL           |
| count     | INT       | true  | NULL           |
| min       | INT       | true  | NULL           |
| median    | INT       | true  | NULL           |
| perc90    | INT       | true  | NULL           |
| perc95    | INT       | true  | NULL           |
| perc96    | INT       | true  | NULL           |
| perc97    | INT       | true  | NULL           |
| perc98    | INT       | true  | NULL           |
| perc99    | INT       | true  | NULL           |
| max       | INT       | true  | NULL           |
| status    | INT       | true  | NULL           |
+-----------+-----------+-------+----------------+
#2

Hey @youngl, thanks for the report. Would you mind pasting the the panic message and stack trace you presumably have as the last lines in the log? The log should be in your <data-dir>/logs.

#3

I don’t see any panic or stack traces in the logs.

Screen Shot 2016-09-29 at 9.46.09 AM.jpg2290×1340 1.26 MB
Screen Shot 2016-09-29 at 9.46.14 AM.jpg2290×1340 1.2 MB
Screen Shot 2016-09-29 at 9.45.56 AM.jpg2290×1340 1.43 MB

#4

Hi! Thanks for reporting this issue to us.
Given the technical nature of this conversation, I would prefer that we continue this troubleshooting session on a GitHub issue. Mind filing an issue for this? https://github.com/cockroachdb/cockroach/issues
Please try to reproduce your issue after starting your server with

cockroach start .... -logtostderr 2>&1 | tee error.log

This will increase the chance that the stack trace / panic message is recorded properly and also save it to a file error.log.

Please also include the output of SHOW CREATE TABLE api; for completeness.

Thanks!

#5

OK, I created the issue: https://github.com/cockroachdb/cockroach/issues/9626

#6

The issue has been fixed ( https://github.com/cockroachdb/cockroach/pull/9629 ) and the fix will be made available in an upcoming beta release. Thanks again for the bug report!