[checkmk-commits] Check_MK Git: check_mk: add bug entry for IPMI agent issue

Module: check_mk Branch: master Commit: e64f82c1e19d15379189671299b12be77af877f5 URL: http://git.mathias-kettner.de/git/?p=check_mk.git;a=commit;h=e64f82c1e19d15… Author: Florian Heigl <fh(a)mathias-kettner.de> Date: Mon Jul 18 15:49:13 2011 +0200 add bug entry for IPMI agent issue --- .bugs/311 | 21 ++++++++++++--------- 1 files changed, 12 insertions(+), 9 deletions(-) diff --git a/.bugs/311 b/.bugs/311 index 9c6b387..e7faede 100644 --- a/.bugs/311 +++ b/.bugs/311 @@ -1,11 +1,14 @@ -Title: manual SNMP checks without scan function not visible in WATO -Component: wato -Benefit: 1 +Title: IPMI cache initialization race condition +Component: checks +Benefit: 6 State: open -Cost: 3 -Date: 2011-06-29 12:56:31 -Class: bug +Cost: 4 +Date: 2011-07-18 15:42:34 +Class: nastiness -When showing the services table of a host in WATO without -Full Scan, then manually configured SNMP services with -out a scan function will not be displayed (e.g. snmp_uptime) +there is a very nasty issue with the IPMI agent code - we +kills the initial build of the SDR cache there will be no point in re-doing the scan and IPMI will be disabled forever. +The problem is that many IPMI modules are not able to handle concurrent requests. If building the SDR cache takes longer than 55+ seconds then we gh chance that a new check_mk_agent call happens via inetd. This second call will cause an IPMI issue which will cause the SDR cache scan to fail. +From that time on, IPMI will be disabled / broken. + +Very ugly issue, maybe xinetd can disable multiple runs of the same service (but this would mean staleness issues will stop the agent forever)