-
Problem report
-
Resolution: Fixed
-
Critical
-
3.4.14
-
Sprint 43, Sprint 44, Sprint 45, Sprint 46, Nov 2018, Sprint 47, Dec 2018, Sprint 48, Jan 2019
-
2
Even after ZBX-6118 fix, there is an issue with updating an item on template, linked to many hosts.
Here is a very slow query in item.update method (new approach implemented in the ZBX-6118, code comment //Check if item with a new key already exists on the child host.):
SQL (365.230574): SELECT i.hostid,i.key_ FROM items i WHERE (i.hostid='26408' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]') OR (i.hostid='23653' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]') OR (i.hostid='12258' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]') OR (i.hostid='30083' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]') OR (i.hostid='27971' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]') OR (i.hostid='25683' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]') OR (i.hostid='26413' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]') OR (i.hostid='16902' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]')...OR (i.hostid='19010' AND i.key_='logrt[/var/log/messages,(error(?!.*dev fd0)),,,skip,,]') LIMIT 1
and possible related issue:
Total time: 949.184311 Total SQL time: 372.596001 SQL count: 6601 (selects: 89 | executes: 6512) Peak memory usage: 4000.75M Memory limit: 16G
~13000 hosts are linked to the template.
There is an opinion from dotneft that, quote: During item update in a template "itemid" order is not ascending. Which lead to possible deadlock, especially with log items.
It's decided to fix it separately of ZBX-3783 because its critical.
- caused by
-
-
- Closed
-
