Zabbix web interface issue with Chrome

[Alexei]

Good news Alexei!

If there is any low level debug information you need, let us I know how to get it for you.

It would be great if you could collect dump of SQL queries executed on DB side when it happens. It would help!

[nelsonab]

As in a show process list or would you like me to see if I can dump the sql queries themselves to disk?

[nelsonab]

Also what about the PHP frontend? Is there a facility for enabeling debug logs? The new 1.8 frontend debug method does not work as the page hangs half way through.

[cronuscronus]

Same issue here

I am seeing the same problem with multiple browsers. This is my current configuration:

Ubuntu 9.10
MySQL 5.1.37 x86_64 (InnoDB)
DM enabled, 1 master node, 4 child nodes, 1 proxy per node (for now)

If I can help by submitting any logs let me know. Here is my processlist output when the frontend is hung:

mysql> show processlist;
+----+--------+------------------------------------------------+-------------+---------+------+---------------+----------------------------------------------------------------------------------------------------+
| Id | User | Host | db | Command | Time | State | Info |
+----+--------+------------------------------------------------+-------------+---------+------+---------------+----------------------------------------------------------------------------------------------------+
| 38 | zabbix | zabbix-frontend:52651 | zabbix_main | Query | 1 | freeing items | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 39 | zabbix | zabbix-frontend:52652 | zabbix_main | Query | 1 | Updating | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 40 | zabbix | zabbix-frontend:52653 | zabbix_main | Query | 2 | freeing items | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 41 | zabbix | zabbix-frontend:52654 | zabbix_main | Query | 1 | Updating | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 42 | zabbix | zabbix-frontend:52655 | zabbix_main | Query | 1 | Updating | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 43 | zabbix | zabbix-frontend:52656 | zabbix_main | Query | 1 | Updating | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 44 | zabbix | zabbix-frontend:52657 | zabbix_main | Query | 1 | Updating | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 45 | zabbix | zabbix-frontend:42236 | zabbix_main | Query | 0 | Updating | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 48 | zabbix | zabbix-frontend:42088 | zabbix_main | Query | 1 | Updating | UPDATE ids SET nextid=nextid+1 WHERE nodeid=1 AND table_name='profiles' AND field_name='profileid' |
| 49 | root | localhost | NULL | Query | 0 | NULL | show processlist |
+----+--------+------------------------------------------------+-------------+---------+------+---------------+----------------------------------------------------------------------------------------------------+

[mschlegel]

I see the same queries as cronuscronus when this shows up on our servers. Those queries are cycling through very quickly as well. Continuing to send the 'show full processlist' command continues to show the same query executing over and over.

Server version: 5.1.41-community-log MySQL Community Server (GPL)
InnoDB with file-per-table.
DM enabled, 1 master node, no active child nodes, 8 proxies
592 hosts, 6320 items, 3881 triggers

[Alexei]

Thank you all for the additional details! It really helps.

[cronuscronus]

Not a problem. We'd like to see the problem solved so let us know what else you need. This seems to only happen on the dashboard. I can cruise around triggers etc... until I'm blue in the face with no problem (all nodes selected).

If it counts for anything I have NoHistorySend=1 and NoEventSend=0 on all nodes.

Thanks,
Aaron

[cronuscronus]

Not a problem. We'd like to see the problem solved so let us know what else you need. This seems to only happen on the dashboard. I can cruise around triggers etc... until I'm blue in the face with no problem (all nodes selected).

If it counts for anything I have NoHistorySend=1 and NoEventSend=0 on all nodes.

Thanks,
Aaron

It looks like this is not a new error:

http://www.zabbix.com/forum/showthread.php?t=6325&page=2

Unfortunately the lock and table format changes haven't worked for me.

[nelsonab]

I just did a fresh install, enabled the zabbix server, opened a new tab and went to "latest data", then opened another tab and pulled up the screens page. The screens page is frozen, I can navigate the latest data page but when I go to click on a graph, no go. Restarting apache is the only fix. :-(

What's interesting however is after reset I can view the screens page just fine. The last time I was able to duplicate this was with a fresh 1.8.1 install.

Here's what I'm seeing:

Every 5.0s: echo "show processlist;" | mysql | grep -v Sleep                  Tue Feb  9 01:57:07 2010

Id      User    Host    db      Command Time    State   Info
214     root    localhost       zabbix  Query   1       freeing items   UPDATE ids SET nextid=nextid+1
 WHERE nodeid=0 AND table_name='profiles' AND field_name='profileid'
215     root    localhost       zabbix  Query   0       Updating        UPDATE ids SET nextid=nextid+1
 WHERE nodeid=0 AND table_name='profiles' AND field_name='profileid'
220     root    localhost       zabbix  Query   1       Updating        UPDATE ids SET nextid=nextid+1
 WHERE nodeid=0 AND table_name='profiles' AND field_name='profileid'
239     root    localhost       zabbix  Query   0       Updating        UPDATE ids SET nextid=nextid+1
 WHERE nodeid=0 AND table_name='profiles' AND field_name='profileid'
344     root    localhost       NULL    Query   0       NULL    show processlist

I watch the info col go from "updating" to "freeing items" but never go away.

[nelsonab]

Again!

I just had it again on my work server. While showing the CEO no less! quite embarrassing.

We were navigating around, going to latest data, then to configuration, then to screens, then pulled up another screen... frozen. Dashboard still worked, but at that point we walked away from the demo.

Mysql still shows the same problem as before.

| 129013 | zabbix | localhost | zabbix | Query   |    0 | Updating      | UPDATE ids SET nextid=nextid+1 WHERE nodeid=0 AND table_name='profiles' AND field_name='profileid' |
| 129016 | zabbix | localhost | zabbix | Query   |    0 | freeing items | UPDATE ids SET nextid=nextid+1 WHERE nodeid=0 AND table_name='profiles' AND field_name='profileid' |
| 129036 | zabbix | localhost | zabbix | Query   |    0 | Updating      | UPDATE ids SET nextid=nextid+1 WHERE nodeid=0 AND table_name='profiles' AND field_name='profileid' |
| 129040 | zabbix | localhost | zabbix | Query   |    0 | Updating      | UPDATE ids SET nextid=nextid+1 WHERE nodeid=0 AND table_name='profiles' AND field_name='profileid' |

[nelsonab]

The common thread for my experiencing this bug seems to be "new". When I do a new install I have a problem, when I add a new user I see it. I'm wondering if there is an issue with the initial setup of a user in the ids table.

[cronuscronus]

The common thread for my experiencing this bug seems to be "new". When I do a new install I have a problem, when I add a new user I see it. I'm wondering if there is an issue with the initial setup of a user in the ids table.

I moved the database to a more powerful server and then changed the db engine on the IDS table to MYISAM and the problem has gone away. Apparently InnoDB requires a lot of horsepower especially when executing so many updates.

[mschlegel]

The question is why some pages are triggering a storm of updates for the ids. My DB servers are not having an issue with the normal query load from the UI or the server, its only this storm of queries that cause an apparent lockup in the UI.

[Aly]

Try this patch, and report if it helps. Patch is against pre 1.8.2 (but may work on 1.8.1 didn't checked)

[wax66]

Excellent, the patch worked for me, thank you! Couldn't patch with patch, though, had to manually add/delete/modify the lines, but now I don't get the database pounding I once had.

Thanks!

Also, here's a patch for the changes I made. It appears the only differences were cosmetic, but it might help someone with 1.8.1 that needs this fix.

-Ron