Hi Jim, Please see my comments inlined with [Nagu].
Thanks -Nagendra High Availability Solutions (www.hasolutions.in) - OpenSAF Support and Services [email protected] Delaware, USA: +1 508-422-7725, India: +91-9866424860 -----Original Message----- From: Carroll, James R [mailto:[email protected]] Sent: 08 October 2018 22:38 To: Nagendra Kumar; [email protected] Subject: RE: EXTERNAL: RE: [users] sending a CER across node boundaries? [follow up] Thanks again for your prompt response. There is one additional item we would like to follow up on: From our own user point of view, this capability is an integral part of the SAF. We don't understand the restriction to local node only - the DN should be valid cluster-wide. Can someone from the community clarify this limitation in regards to the following: 1) Is this a feature that is typically not used, across nodes, by the average user? Maybe we are misunderstanding how to use it? If not, is there some other recommendation to handle the reporting of an error on another node. [Nagu]: In general, the applications reports its error itself or the proxy/wrapper reports errors of its proxied components. Also, proxy/wrapper resides at the same node. (Internode proxy-proxied component is not supported). Also, if there is any failure in component, then any way, Amf detects it locally, so it doesn't require error reporting. Local error reporting or local error detection is always faster, so user prefer that case. But of course, you can have one admin app, which can monitors the apps at all the nodes and report error when detected. If that is the case, then few alternates for you could be: - you should issue admin command 'component/su restart' on the components/su/node from that admin application. Admin commands can be issued from any node for any component/su(because imm data is the same on all the nodes). Please note that if the application has other recovery configured then this option is not appropriate. Also, it will delay in the service recovery. - you should be able to issue kill command (from admin app) on the faulty app on the remote node (by login into the remote system) and once the applications is killed, let Amf take actions. - you should be able to login and run a script on the remote node, which will spawn a process to report error to Amf(There is disadvantage of this approach as spawning a new process will take some time and will delay the fault detection and recovery). Can you please provide more information on your use case? 2) is there an inherently large amount of complexity involved once you go off the node? [Nagu]: As per OpenSAF architecture, Directors(residing at controllers) contains Cluster wide information and all the node directors(residing at each node) have the node specific information. Directors provide only those information to node directors, which are configured for that node(like SU, Components). This is the primarily reason that if a component1(on Node1) reports error on a component2 (residing on Node2), then node director(which interact with components directly) of Node1 doesn't recognize Component2 as it has information about Component and SUs of Node1 only. So, it returns Error-Not-Exist. Hence, it has moderate complexity to accept the Error Report for Component residing at another node, because it need to send that message to Director to verify whether this component is configured or not and then after verification, it need to send it to the node director to the concerned node, where this component belongs to(for taking actions). Thanks. Jim -----Original Message----- From: Nagendra Kumar <[email protected]> Sent: Wednesday, October 03, 2018 1:36 PM To: Carroll, James R (US) <[email protected]>; [email protected] Subject: RE: EXTERNAL: RE: [users] sending a CER across node boundaries? Hi Jim, It is still an existing limitation. Thanks, Nagendra, 91-9866424860 High Availability Solutions (www.hasolutions.in) - OpenSAF Support and Services Delaware, USA. Hyderabad, India. -----Original Message----- From: Carroll, James R [mailto:[email protected]] Sent: 03 October 2018 22:46 To: Nagendra Kumar; [email protected] Subject: RE: EXTERNAL: RE: [users] sending a CER across node boundaries? Thank you for your response. Do you know if this limitation is resolved in a more recent version of OpenSAF, or if this is still an existing limitation? Jim -----Original Message----- From: Nagendra Kumar <[email protected]> Sent: Wednesday, October 03, 2018 1:03 PM To: Carroll, James R (US) <[email protected]>; [email protected] Subject: EXTERNAL: RE: [users] sending a CER across node boundaries? Hi James, >> Is there any know limitation in OpenSAF, that prevents sending a CER across node boundaries? Yes, it is documented in Amf Programmer's Reference document. 2.2.2 Compliance Report: Table 1 Compliance Table - Availability Management Framework(SAI-AIS-AMF-B.04.01) 7.12.1 saAmfComponentErrorReport_4() Yes Partly "Only supported value for rootCorrelationId and parentCorrelationId is: SA_NTF_IDENTIFIER_UNUSED. Only supported for local node erroneous component, otherwise return ERR_NOT_EXIST." Thanks, Nagendra, 91-9866424860 High Availability Solutions Pvt. Ltd. (www.hasolutions.in) - OpenSAF Support and Services -----Original Message----- From: Carroll, James R [mailto:[email protected]] Sent: 03 October 2018 21:23 To: [email protected] Subject: [users] sending a CER across node boundaries? Hi all, I am using OpenSAF 5.2, and I am running into an issue when trying to issue a Component Error Report (CER) from Node A, against a component on Node B. Both nodes are in the same cluster, and we have confirmed the DN, but when we issue the CER, we are getting the error code=12 SA_AIS_ERR_NOT_EXIST. Is there any know limitation in OpenSAF, that prevents sending a CER across node boundaries? Thanks. Jim _______________________________________________ Opensaf-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/opensaf-users _______________________________________________ Opensaf-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/opensaf-users
