[lvs-users] IPVS error : "no destination available"
khosrow.ebrahimpour at ec.gc.ca
Tue Sep 18 00:34:55 BST 2012
On 2012-09-14, at 5:12 PM, Khosrow Ebrahimpour wrote:
> Hi List,
> I run a cluster of ldap servers behind LVS and recently started to see
> connection timeouts from clients that are connected to the loadbalancer. If
> those same clients are connected directly to the realserver, there is no
> The loadbalancer is running LVS-DR with a LC scheduler with two realservers.
> And each time there's a timeout error both realservers are available and
> there's no error logged by the LDAP service.
> After looking through various log files, I found that about once a day I get
> this error:
> Sep 13 06:27:20 lvs1 kernel: [39885980.112929] IPVS: lc: TCP 10.10.10.11:389 -
> no destination available
> I was unable to find any kernel variables to increase max connections (I looked
> here http://www.kernel.org/doc/Documentation/networking/ipvs-sysctl.txt).
> A thread on the lvs-devel ( http://archive.linuxvirtualserver.org/html/lvs-
> ) list mentions the variable ip_vs_dest_totalconns. Is there a way to
> increase this?
> Any help is appreciated.
I never set the "--l-threshold" and "--u-threshold" variables and they are both set to 0. And so if my understanding of IPVS is correct, I should never hit that limit. Here's a quick snapshot of my thresholds and the connections as they are right now.
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Uthreshold Lthreshold ActiveConn InActConn
TCP ldap.example.com:ldap lc
-> ldap1.example.com:ldap 0 0 159 46
-> ldap2.example.com:ldap 0 0 156 297
TCP ldap.example.com:ldaps lc
-> ldap1.example.com:ldaps 0 0 601 1065
-> ldap2.example.com:ldaps 0 0 607 138
As you can see the systems are busy, but don't seem to be overloaded in any way.
More information about the lvs-users