dc_errors ( and possibly others) data collector table stops retaining data after a while.
It appears that dc_errors is empty despite a retention policy of 1000KB memory, 10GB(~) disk,
I noticed this on 2 different clusters running 7.1.1-x versions . The collection process restart after rerunning the
SELECT SET_DATA_COLLECTOR_POLICY('Errors', '2000', '1000000'); with the same parameter.
BEFORE
========
dbadmin=> SELECT GET_DATA_COLLECTOR_POLICY('Errors');
GET_DATA_COLLECTOR_POLICY
---------------------------------------------------------------
1000KB kept in memory, 10000000KB kept on disk. kept on disk.
(1 row)
Time: First fetch (1 row): 86.923 ms. All rows formatted: 86.985 ms
dbadmin=> select count(*) from dc_errors;
count
-------
0
(1 row)
Time: First fetch (1 row): 291.413 ms. All rows formatted: 291.498 ms
dbadmin=> select version();
version
-------------------------------------
Vertica Analytic Database v7.1.1-10
(1 row)
AFTER
======
SET_DATA_COLLECTOR_POLICY
---------------------------
SET
(1 row)
Time: First fetch (1 row): 586.010 ms. All rows formatted: 586.971 ms
dbadmin=> select count(*) from dc_errors;
count
-------
1
(1 row)
The data collector table shows no records and a lot of retired files. This behavior is not according to data collector design in my opinion. Is this a bug in version 7 ?
Comments
Hi colin_loghin,
We do have a known issue with retention policy management in 7.1.1 that is scheduled to be resolved in 7.1.2 which we hope will be available shortly.
We just released 7.1SP2 which should have this issue resolved!
See https://community.dev.hp.com/t5/Vertica-Forum/HP-Vertica-7-1-SP2-is-now-available/m-p/230003#M10197