Hi,
In our 1 Master - several slaves environment master slapd sometimes stucks in such condition that it takes minutes to get replay to simple queries with no visible resource consumption on both: the slapd process itself and the linux it is running on.
Restarting slapd helps for several days. Today I have not been able to restart it without kill -9 because of:
slapd shutdown: waiting for 12 operations/tasks to finish
Waited 5 minutes with no luck, kill -9, start.
slapd: 2.4.28 OS: Ubuntu 12.04.2 LTS
I'm Ok with "try latest version", but it would be nice to have some pointers (maybe bug in this version or error in configuration).
The server is running for 6+ month with no problems. The stucks first appeared 2 weeks ago after changing ACLs to use group.exact instead of set which came deadly slow.
--On November 19, 2013 at 8:20:57 PM +0200 Покотиленко Костик casper@meteor.dp.ua wrote:
Hi,
In our 1 Master - several slaves environment master slapd sometimes stucks in such condition that it takes minutes to get replay to simple queries with no visible resource consumption on both: the slapd process itself and the linux it is running on.
Restarting slapd helps for several days. Today I have not been able to restart it without kill -9 because of:
slapd shutdown: waiting for 12 operations/tasks to finish
Waited 5 minutes with no luck, kill -9, start.
slapd: 2.4.28 OS: Ubuntu 12.04.2 LTS
I'm Ok with "try latest version", but it would be nice to have some pointers (maybe bug in this version or error in configuration).
I would definitely say upgrade to 2.4.38. One thing that particularly stands out from the Changes file:
OpenLDAP 2.4.32 Release (2012/07/31) Fixed slapd-bdb/hdb cache hang under high load (ITS#7222)
Generally, I would suggest getting a full gdb backtrace and looking at the output of db_stat -C A and db_stat -c when it is in this state, if it occurs again after running a current release.
--Quanah
В Вто, 19/11/2013 в 11:11 -0800, Quanah Gibson-Mount пишет:
--On November 19, 2013 at 8:20:57 PM +0200 Покотиленко Костик casper@meteor.dp.ua wrote:
Hi,
In our 1 Master - several slaves environment master slapd sometimes stucks in such condition that it takes minutes to get replay to simple queries with no visible resource consumption on both: the slapd process itself and the linux it is running on.
Restarting slapd helps for several days. Today I have not been able to restart it without kill -9 because of:
slapd shutdown: waiting for 12 operations/tasks to finish
Waited 5 minutes with no luck, kill -9, start.
slapd: 2.4.28 OS: Ubuntu 12.04.2 LTS
I'm Ok with "try latest version", but it would be nice to have some pointers (maybe bug in this version or error in configuration).
I would definitely say upgrade to 2.4.38. One thing that particularly stands out from the Changes file:
OpenLDAP 2.4.32 Release (2012/07/31) Fixed slapd-bdb/hdb cache hang under high load (ITS#7222)
Generally, I would suggest getting a full gdb backtrace and looking at the output of db_stat -C A and db_stat -c when it is in this state, if it occurs again after running a current release.
Thanks.
openldap-technical@openldap.org
-
Quanah Gibson-Mount -
Покотиленко Костик