Please see inline. Thanks, Praveen On 23-Jun-14 10:14 PM, Mark Garstin wrote: > I have a Proxy/Proxied configuration that I'm running integration testing on. > I have two tests whereby the Proxy is terminated. One of them behaves as > expected but the other one doesn't. Here are the cases. > > > 1. If I do a kill -9 on the Proxy process it goes away but is > immediately restarted by AMF. My Proxy then reconnects with the Proxied > component (which has been running all along) and everyone is fine. > > This test case is good. > > 2. If I enter on the command line amf-adm lock <Proxy's DN> followed by > amf-adm lock-in <Proxy's DN> the Proxy is terminated gracefully but then a > few seconds later (7 seconds to be exact) I get an error log from AMF stating > : > > osafamfnd[3470]: NO > 'safComp=MyProxiedComp,safSu=SU1,safSg=MyProxiedSG,safApp=MyApp' faulted due > to 'prxRegTimeout' : Recovery is 'suFailover' > osafamfnd[3470]: NO 'safSu=SU1,safSg=MyProxiedSG,safApp=MyApp' Presence State > INSTANTIATED => TERMINATING > > Followed by a cleanup of the Proxied Component. A proxied actually moves to orphan state. Please see sample notifications thant AMF sends when proxied becomes orphans in #108. In orphan state if no other proxy component registers in within saAmfCtDefClcCliTimeout, than AMF reports error on the proxied component. In your application, proxy SU is locked and locked-in then some other process should register the proxy. Please check also in the application proxied instantiate timeout should be 7 seconds.
> Now, according to the AMF B.04.01 spec, section 5.3, page 219, the Proxied > component should enter the UNPROXIED state: > > "If a proxy component fails, the Availability Management Framework may > perform a > fail-over if fail-over is allowed by the redundancy model of the service > group to which > the proxy belongs. During the proxy component fail-over procedure, the > Availability > Management Framework implicitly unregisters all registered proxied components > associated with the failing proxy component. However, this implicit > unregistration > should not be considered by the Availability Management Framework as a sign > of a > proxied component failure. The implicit unregistration simply indicates that > the proxy > component is unable to continue the "proxying" work, and the Availability > Management > Framework should find another proxy component to take over the "proxying" > work." > > A bit further on in that section it states: > > "If the Availability Management Framework is unable to find another proxy > component > to "proxy" a given proxied component, the given proxied component shall enter > the > SA_AMF_PROXY_STATUS_UNPROXIED status (defined in Section 7.4.4.8), and an > appropriate alarm shall be issued by the Availability Management Framework > (see > Section 11.2.1.5 on page 427) to indicate this situation. Whenever a proxied > component > enters the SA_AMF_PROXY_STATUS_PROXIED status (defined in > Section 7.4.4.8), an appropriate notification (see Section 11.2.2.7 on page > 434) will > be issued to indicate the change in the status of the proxied component." > > My interpretation of these paragraphs (especially the parts highlighted in > red) is that the Proxied component should never be terminated, but it is in > this scenario. > > Now, perhaps this has to do with how I've configured my Proxy and Proxied > components. I have 2 redundant nodes running these Proxy/Proxied component > pairs. The Proxy's on both nodes are members of separate SG's that are > configured as no redundancy (so both Proxy's are ACTIVE all the time). But > the two Proxied components are members of just one SG which is configured for > 2N redundancy (so one Proxied component is ACTIVE and the other is STANDBY). Since proxy is part of No Red, when proxy SU is locked its SI becomes unassigned if there is no spare SU. So if there is some spare SU hosted on the same node, AMF will assign this SI to the spare the spare SU. This spare SU can register also register proxied components again. But all this should happen within limit of saAmfCtDefClcCliTimeout. > If what I'm doing is correct but OpenSAF is reacting incorrectly then perhaps > a ticket should be raised on this issue. If this has already been addressed > then please tell me in which release it has been corrected (4.4 or 4.5). If, > however, I have configured the SG's of the components incorrectly or I'm > interpreting the spec incorrectly then please, someone slap me up the side of > my head and show me the error of my ways. > > Cheers, > > > > Mark Garstin > Sr. SW Developer > Ericsson > > ------------------------------------------------------------------------------ > HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions > Find What Matters Most in Your Big Data with HPCC Systems > Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. > Leverages Graph Analysis for Fast Processing & Easy Data Exploration > http://p.sf.net/sfu/hpccsystems > _______________________________________________ > Opensaf-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/opensaf-users ------------------------------------------------------------------------------ Open source business process management suite built on Java and Eclipse Turn processes into business applications with Bonita BPM Community Edition Quickly connect people, data, and systems into organized workflows Winner of BOSSIE, CODIE, OW2 and Gartner awards http://p.sf.net/sfu/Bonitasoft _______________________________________________ Opensaf-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/opensaf-users
