[ZBX-14624] Agent collector not releasing connection to missing/removed devices and gets full Created: 2018 Jul 20  Updated: 2024 Apr 10  Resolved: 2019 Jan 28

Status: Closed
Project: ZABBIX BUGS AND ISSUES
Component/s: Agent (G)
Affects Version/s: None
Fix Version/s: 3.0.25rc1, 4.0.4rc1, 4.2.0alpha3, 4.2 (plan)

Type: Problem report Priority: Minor
Reporter: Edmunds Vesmanis Assignee: Martins Valkovskis
Resolution: Fixed Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Duplicate
Team: Team D
Team: Team D
Sprint: Sprint 38, Sprint 39, Sprint 40, Sprint 41, Sprint 42, Sprint 43, Sprint 44, Sprint 45, Sprint 46, Nov 2018, Sprint 47, Dec 2018, Sprint 48, Jan 2019
Story Points: 2

 Description   

If system has devices with dynamically changing paths or if device gets manually removed
then agent won't release it and will keep connection resulting in overfilling.
Error message also shown in frontend:

# /opt/zabbix/bin/zabbix_get -s 127.0.0.1 -k vfs.dev.read[/dev/sdXX##,ops,avg1]
ZBX_NOTSUPPORTED: Cannot add disk device to agent collector.

In debug mode 5:

    659 35331:20180720:103340.955 In collector_diskdevice_add() devname:'sdXX##'
    660 35331:20180720:103340.955 collector_diskdevice_add() collector is full

 Comments   
Comment by richlv [ 2018 Jul 20 ]

Might the implementation from https://support.zabbix.com/browse/ZBXNEXT-801?focusedCommentId=58113&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-58113 .

Comment by Jurijs Klopovskis [ 2019 Jan 02 ]

Fixed in 3.0.25rc1 (r88176, r88179), 4.0.4rc1 (r88177, r88181, r88182) and 4.2 / trunk (r88178).

Comment by Alexander Vladishev [ 2019 Jan 22 ]

Updated documentation:

Comment by richlv [ 2019 Jan 22 ]

Thank you for the fix and great documentation updates. While it might be a bit too much for the documentation, maybe an example scenario can be shared here, in which case items would return 0?

Comment by Jurijs Klopovskis [ 2019 Jan 22 ]

richlv, it's pretty simple.
If vfs.dev.read or vfs.dev.write items are used with delay values >= 3 hours, you will always get 0. That's it.

 

Comment by richlv [ 2019 Jan 24 ]

Thanks, I was expecting it to be a bit more of an edge case. I'd suggest to document it also next to each affected item, as the small note is very easy to overlook, and not that easy to connect with the affected item keys.

Comment by Alexander Vladishev [ 2019 Jan 25 ]

I agree with richlv. It should be made same way as in "Supported macros" section.

Comment by Martins Valkovskis [ 2019 Jan 25 ]

It is now also briefly mentioned with the items and linked to the footnotes containing the full note.

Generated at Fri Apr 19 16:47:09 EEST 2024 using Jira 9.12.4#9120004-sha1:625303b708afdb767e17cb2838290c41888e9ff0.