Cookie invalidation after quiesce
Hi all,
I'm preparing HA installation of Pega 7.4 with cookie invalidation. Pega is running on WebSphere 9. The channel setting CookiesConfigureNoCache is False. HAProxy is used for cookie insertion with backend server name.
prconfig.xml contains the following lines:
<env name="session/ha/quiesce/customSessionInvalidationMethod" value="configurable-cookie-invalidator" /><env name="session/ha/quiesce/cookieToInvalidate" value="ha_session" /><env name="session/ha/Enabled" value="true" />
Accepted Solution
Pegasystems Inc.
US
Upon reviewing the associated Support Request, this was the resolution:
Perform the following local-change: *
Please make sure that "initialization/settingsource" is set to "merged" in the prconfig for each node (7.4 should have this set by default, but in case you are copying a prconfig from an older system you will need this).
And then set the following DSS + restart all nodes:
name: prconfig/timeout/page/default
owning ruleset : Pega-Engine
value: some number
name: prconfig/timeout/browser/default
owning ruleset : Pega-Engine
value: some number
The values to DSS should be set to communicated defaults and then corrected based on your observations.
Pegasystems Inc.
US
As of Pega 7.2.2, the default quiesce strategy is immediate drain. Based on the configuration shown above, you are running with the default strategy. Immediate drain quiesce no longer relies on cookie invalidation when quiesce is initiated. Subsequent requests will be directed to the node previously handling the session based on the cookie until the node is removed from the load balancer.
Sberbank
RU
Thanks for the clarification! So passivation with immediate drain should be simple: quiesce the node and remove it from the load balancer. That's clear statement I missed in the documentation.
I tried slow drain mode too but browser did not receive clear cookie header. The configuration was the same but with additional line:
<env name="session/ha/quiesce/strategy" value="slowDrain" />
Is there anything else I should reconfigure for slow drain mode?
Pegasystems Inc.
US
Your configuration for slow drain quiesce looks complete. Please confirm that the node was restarted after the configuration change was made and that slow drain quiesce is in use when the quiesce process is initiated.
The following log will include the active quiesce strategy:
2018-10-26 10:22:07,124 [b0f22014a01f.event-5] [ STANDARD] [ ] [ ] ( pega.highavailability.quiesce) INFO - [BeginQuiesce] This node has started the quiesce process. Mode: immediateDrain
Sberbank
RU
Yes, the node had been restarted before I tried that. And the log confirms it:
2018-11-13 17:21:40,777 [ WebContainer : 6] [STANDARD] [PegaRULES:07.10] (com.pega.highavailability.quiesce) INFO vck3-s528-was001-was-node-2| Proprietary information hidden|SOAP|PegaAESRemote|RuleData|ExecuteMBeanOperation|AA1W70VQVPRC5K0QRJR0MVY883R2OM9SV - [BeginQuiesce] This node has started the quiesce process. Mode: slowDrain Requestor ID: AA1W70VQVPRC5K0QRJR0MVY883R2OM9SV
Accepted Solution
Pegasystems Inc.
US
Upon reviewing the associated Support Request, this was the resolution:
Perform the following local-change: *
Please make sure that "initialization/settingsource" is set to "merged" in the prconfig for each node (7.4 should have this set by default, but in case you are copying a prconfig from an older system you will need this).
And then set the following DSS + restart all nodes:
name: prconfig/timeout/page/default
owning ruleset : Pega-Engine
value: some number
name: prconfig/timeout/browser/default
owning ruleset : Pega-Engine
value: some number
The values to DSS should be set to communicated defaults and then corrected based on your observations.