[ZBX-9733] Zabbix agent port was taken by SYSTEM user and running without process on top of it. Created: 2015 Jul 27 Updated: 2017 May 30 Resolved: 2016 Jan 25 |
|
| Status: | Closed |
| Project: | ZABBIX BUGS AND ISSUES |
| Component/s: | Agent (G) |
| Affects Version/s: | 2.4.1 |
| Fix Version/s: | 2.2.12rc1, 2.4.8rc1, 3.0.0beta2 |
| Type: | Incident report | Priority: | Major |
| Reporter: | Natalia Kagan | Assignee: | Unassigned |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | windows | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Environment: |
zabbix server & proxy 2.4.4 on centos 6, DB: mysql |
||
| Issue Links: |
|
||||
| Description |
|
We encountered to some issue when the zabbix port 10050 was up and running twice, once with system user, and another was the agent: C:\Users\XXXX>netstat -aon | find "10050" TCP 0.0.0.0:10050 0.0.0.0:0 LISTENING 16496 TCP 0.0.0.0:10050 0.0.0.0:0 LISTENING 12600 TCP [::]:10050 [::]:0 LISTENING 12600 TCP [::]:10050 [::]:0 LISTENING 16496 while proccess: 16496 is owned by zabbix_agent and process: 12600 was not in task manager list (which means that the process is no longer exist). This prevents our zabbix proxy not getting response to non active checks from the host: "Get value from agent failed: cannot connect to [[PPP.YYY.XXX.ZZZ]:10050]: [111] Connection refused" We were not able to kill the process and also not able to terminate the port since the process wasn't exist (tried with TCPView too). We also found bunch of Time_wait connections on port 10050 which were terminated by TCP Viewer, but the port LISTEN operation wasn't resolved. Only reboot solved the issue, but in production we can't reboot around 50 servers which running on them critical services. In addition we don't see this phenomenon on all windows machines. Do you have any idea why it's happening? Thanks, |
| Comments |
| Comment by Aleksandrs Saveljevs [ 2015 Jul 28 ] |
|
Unless you have a reliable way of reproducing the issue, it looks more like a support request, rather than a bug report. For support, https://www.zabbix.org/wiki/Getting_help is a better place. |
| Comment by Aleksandrs Saveljevs [ 2015 Jul 28 ] |
|
Actually, I have just tried starting two Windows agents with the same configuration file and... it worked and they listened to the same port. It should not, so let's keep this issue open. |
| Comment by Oleksii Zagorskyi [ 2015 Aug 20 ] |
|
Something similar was discussed some time ago in |
| Comment by Aleksandrs Saveljevs [ 2016 Jan 20 ] |
|
According to the following links, SO_REUSEADDR has a different semantics on Windows than on Unix, and it is suggested to use SO_EXCLUSIVEADDRUSE instead:
The only problem is that it seems to require administrative privileges on Windows XP and earlier, but then setsockopt() call will simply fail silently - the agent will still work. |
| Comment by Aleksandrs Saveljevs [ 2016 Jan 20 ] |
|
Fixed in development branch svn://svn.zabbix.com/branches/dev/ZBX-9733 . |
| Comment by Andris Zeila [ 2016 Jan 22 ] |
|
Successfully tested |
| Comment by Aleksandrs Saveljevs [ 2016 Jan 22 ] |
|
Fixed in pre-2.2.12rc1 r57930, pre-2.4.8rc1 r57931, pre-3.0.0beta2 (trunk) r57932. |
| Comment by Aleksandrs Saveljevs [ 2016 Jan 22 ] |
|
(1) There were conflicts merging into trunk and it took me a while to fix it. Please review r57932, r57934, r57935. wiper CLOSED |