View Full Version : is it possible to flush the queue
When a host is not reachable..the queue gets filled up and slowes down the zabbix to a point that it would not collect new data.
Is it possible to flush the queue in 1.1b5 version?
KarmaPolice
03-02-2006, 23:56
same problem here... looking for answer.
ci_newman
05-07-2006, 11:51
Im currently using the 1.1 release version, and Im having issues with all of the items in the queue sitting at +5 minutes (The re-check date is set to 1970)
Did anyone ever find a solution to this? I want to try and flush the queue to see if it'll reset everything.
Nate Bell
05-07-2006, 15:51
I poked around the Zabbix DB and didn't find an easy way to flush the queue. The way the queue is formed is by reading the timestamp for each item's next time to be updated. If that time is coming up soon, it gets listed in the queue.
I tried tweaking these values by hand (advancing the timestamp forward in time), but I think I ended up doing more damage than good. Flushing the queue is one of those requests that has been around for a while, but I'm not sure there's a standard idea of what that means.
I think flushing would mean advancing the timestamp forwards for blocking items so other items get a chance to move through the queue. One way to do this would be to allow you to choose which items you want to flush and clicking a "Flush" button which sets the next update timestamp for those items to "now" (or possibly "now"+"their update interval"). I think that would let the blocked items pass through the queue. That's just my initial thought. I'm sure someone else can come up with something better.
Nate
dantheman
05-07-2006, 16:28
I had that problem, and in my case, all items with a 1970's date were all active checks and the server that was having this problem, I had an incorrect hostname set in that server's agent config file... so maybe you should check that out, see if that is what's causign the problem.
I had the same problem after having my server down due to a mysql server migration. The only way I was able to fix it was to run an update query and set the nextcheck value to UNIX_TIMESTAMP(). After doing that, Zabbix functioned normally.
ad@kbc-clearing.com
24-08-2006, 22:40
Hi there,
Just went live with 1.1 stable release.
Experiencing the same problem ; if one host (out of hundred) is not reachable, then the queue fills up and many items are not collected any more.
Anyone found a solution for this problem ?
Is Alexei aware of this isue ?
Ad
disgruntleddutch
08-11-2006, 00:08
Same happens here too. Spent hours trying to tweak MySQL but the queue just fills up. I ended up changing everything checking under 60 seconds to 60 seconds and others to agent (active).
Note that ZABBIX 1.1.3 has better and more understandable view of the Queue. Previous releases incorrectly calculated active items in the queue.
I came across the same issue with version 1.1.3, 2 items in the queue were added (not sure how or why) with the 10AM 1970 (unix zero time) entries, then the > 5 minute queue number kept growing.
This is happening as we speak, I'm yet to find a resolution to it.
Nate Bell
10-11-2006, 15:46
Our Zabbix server has been suffering more and more from this problem. It's to the point where a host can crash, which causes the queue to backlog, and we don't hear about it from Zabbix because the backlogged queue prevents Zabbix from reporting anything.
This situation happens either when a host crashes, or all its CPU cycles are devoted to a task. Either way, every item gets a gap in its history until the problem is resolved. I have the feeling Zabbix is getting stuck when it tries to collect a UserParameter on a downed/busy host. The servers that seem to make the queue backlog are servers with a handful of UserParameters. Hosts with fewer UserParameters don't seem to be effected.
I'm currently cleaning up our installation of Zabbix by removing as many UserParameters as I can, and stripping out unused items. I'm hoping to return our Zabbix installation to its former glory. I'll report on how sucessful I am.
Nate
I came across the same issue with version 1.1.3, 2 items in the queue were added (not sure how or why) with the 10AM 1970 (unix zero time) entries, then the > 5 minute queue number kept growing.
This is happening as we speak, I'm yet to find a resolution to it.
This issue seemed to have resolved itself on our setup (11 hosts, more to come). To help this issue along, we increased the number of zabbix_server pollers to 15, restarted zabbix_server and also restarted the zabbix_agentd on the host which was affecting the queue.