Hello I have a corosync-pacemaker Cluster with two nodes (mutual takeover). All resource locations on one node depending on the nodes network connectivitiy. pingd-depending rules for the ip interfaces in a base - group and then colocations for all other groups. I wrote rules around pingd for the following situations:
Notation: - Node a - Node b - "ip a" is the ip interface resource on node a - "ip b" is the ip interface resource on node b - pingd resource for a, running on node a (with score infinity), host to ping is the default gateway, dampen 60s, operations interval 5s, operations timeout 25s - pingd resource for b, running on node b (with score infinity), host to ping is the default gateway, dampen 60s, operations interval 5s, operations timeout 25s (both nodes are in one subnet with one gateway) The Rules: base location - score for ip a on node a = 500 base location - score for ip b on node b = 500 if node a can ping the defaultgateway then score for ip a on node a = +1500 if node b can ping the defaultgateway then score for ip a on node b = +1000 if node b can't ping defaultgateway then score for ip a on node b = -INFINITY if node b can ping the defaultgateway then score for ip b on node b = +1500 if node a can ping the defaultgateway then score for ip b on node a = +1000 if node a can't ping defaultgateway then score for ip b on node a = -INFINITY What I want: If one node can't reach the defaultgateway, all resources running on that node, should switch to the other node, IF the other node can still reach the defaultgateway. If the "other" node also can't reach the defaultgateway, nothing should happen and all resources should stay on the node they are currently running. The Problem: If I unplug the network cable on both nodes (at one time), the resources should stay on the node they are currently running. BUT one node detects the failed gateway ping earlyer, so the cluster trys to switch its resources. After a short time, the other node detects the failed gateway and all switched resources goes back to the node they running before. I thought the attribute "dumpen" is the solution to handle this problem? All other cases are working fine. The Config (constraints section only): <constraints> <rsc_location id="location_dbma03_gateway_reachable" node="dbma03" rsc="resource_pingd_dbma03_gateway_reachable" score="+INFINITY"/> <rsc_location id="location_dbma04_gateway_reachable" node="dbma04" rsc="resource_pingd_dbma04_gateway_reachable" score="+INFINITY"/> <rsc_location id="dbma03_base_group_default" node="dbma03" rsc="resource_dbma03s_ip" score="500"/> <rsc_location id="dbma04_base_group_default" node="dbma04" rsc="resource_dbma04s_ip" score="500"/> <rsc_location id="location_resource_dbma03s_ip_ping_depending" rsc="resource_dbma03s_ip"> <rule id="rule-resource_dbma03s_ip-standard-location" score="+1500"> <expression attribute="dbma03_gateway_reachable" id="ping-resource_dbma03s_ip-03-normal-state-rule-condition1" operation="eq" value="100"/> <expression attribute="#uname" id="ping-resource_dbma03s_ip-03-normal-state-rule-condition2" operation="eq" value="dbma03"/> </rule> <rule id="ping-resource_dbma04s_ip-03-normal-state" score="+1000"> <expression attribute="dbma04_gateway_reachable" id="ping-resource_dbma04s_ip-03-normal-state-rule-condition1" operation="eq" value="100"/> <expression attribute="#uname" id="ping-resource_dbma04s_ip-03-normal-state-rule-condition2" operation="eq" value="dbma04"/> </rule> <rule id="xping-resource_dbma04s_ip-03-normal-state" score="-INFINITY"> <expression attribute="dbma04_gateway_reachable" id="xping-resource_dbma04s_ip-03-normal-state-rule-condition1" operation="lt" value="100"/> <expression attribute="#uname" id="xping-resource_dbma04s_ip-03-normal-state-rule-condition2" operation="eq" value="dbma04"/> </rule> </rsc_location> <rsc_location id="location_resource_dbma04s_ip_ping_depending" rsc="resource_dbma04s_ip"> <rule id="rule-resource_dbma04s_ip-standard-location" score="+1500"> <expression attribute="dbma04_gateway_reachable" id="ping-resource_dbma04s_ip-04-normal-state-rule-condition1" operation="eq" value="100"/> <expression attribute="#uname" id="ping-resource_dbma04s_ip-04-normal-state-rule-condition2" operation="eq" value="dbma04"/> </rule> <rule id="ping-resource_dbma03s_ip-04-normal-state" score="+1000"> <expression attribute="dbma03_gateway_reachable" id="ping-resource_dbma03s_ip-04-normal-state-rule-condition1" operation="eq" value="100"/> <expression attribute="#uname" id="ping-resource_dbma03s_ip-04-normal-state-rule-condition2" operation="eq" value="dbma03"/> </rule> <rule id="Xping-resource_dbma03s_ip-04-normal-state" score="-INFINITY"> <expression attribute="dbma03_gateway_reachable" id="Xping-resource_dbma03s_ip-04-normal-state-rule-condition1" operation="lt" value="100"/> <expression attribute="#uname" id="Xping-resource_dbma03s_ip-04-normal-state-rule-condition2" operation="eq" value="dbma03"/> </rule> </rsc_location> </constraints> Thanks Heinz <https://img.web.de/p.gif> WEB.DE DSL: Internet, Telefon und Entertainment für nur 19,99 EUR/mtl.! http://produkte.web.de/go/02/
<<p.gif>>
_______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
