Distributed monitoring and update from 1.8 to 2.0.3

[soulhunter]

Greetings,

I recently updated our small DM setup from 1.8 series to 2.0.3 (PostgreSQL database). .sql patch applied cleanly, and monitoring seems to be working just fine: graphs, alerts, ....

Now, my only concern is that I'm seeing this on zabbix-server logs, this is on child node:

22956:20121016:093911.303 [Z3005] query failed: [0] PGRES_FATAL_ERROR:ERROR: column "slideshowid" does not exist
LINE 1: select slideshowid,screenid,step,delay,httptestid,name from ...
^
[select slideshowid,screenid,step,delay,httptestid,name from hosts where hostid=200100000000001]
(.....)
9663:20121016:094314.244 [Z3005] query failed: [0] PGRES_FATAL_ERROR:ERROR: column "macro" does not exist
LINE 1: select macro from graphs_items where gitemid=200200000000684
^
[select macro from graphs_items where gitemid=200200000000684]
(...)

And similar ones on master node:

22956:20121016:094925.213 [Z3005] query failed: [0] PGRES_FATAL_ERROR:ERROR: column "hostid" does not exist
LINE 1: select hostid from graphs_items where gitemid=20020000000079...
^
[select hostid from graphs_items where gitemid=200200000000793]
(.....)
22957:20121016:095135.625 NODE 1: Received history_uint from node 2 for node 2 datalen 1107
22956:20121016:095142.226 cannot find table [hosts_profiles]

And so on. I think I tracked this to node_cksum table. Maybe this is a leftover of a couple of nodes that were slightly out of sync before starting update procedure, but not sure on that.

Any ideas?

Ildefonso.

[ebuck]

I'm seeing the same thing. Upgraded Zabbix 1.8.10 to 2.0.3 (MySQL database, two node DM). I have two errors:

query failed: [1054] Unknown column 'internal' in 'field list' 
[select name,internal,name,hsize,vsize,templateid from hosts where hostid=200100000000001]

query failed: [1054] Unknown column 'macro' in 'field list' 
[select macro from graphs_items where gitemid=200200000000047]

Did you resolve this? Is it safe to truncate the node_cksum table?

[soulhunter]

Well, because my configuration was relatively small, and I was not using the distributed monitoring that much (I just wanted to use it to "be ready" in case we had to install on a larger environment, with some special access requirements: that have not happened yet), I decided to just drop current DB, and re-create it (yeah, sounds lazy, but I just got tired of the message on the logs, and really: it was fast to re-add the hosts, also, I saw that the newer templates, included with Zabbix 2.0 were interesting, so, decided to have a "fresh start" with 2.0).

Without more understanding, I can't say for sure if you can or can't truncate node_cksum table, I feel it should be safe to delete the offending rows, but I am not sure on how to identify them. It would be good if someone with more in-deep knowledge help us here.

Ildefonso.

[ebuck]

I went ahead and stopped both the master and slave nodes and deleted all rows from the node_cksum table on both. After restarting the master then slave, the tables were re-populated. The errors are now gone and there don't appear to be any negative effects.