Hi Girish, Yes, it reads from local replica in case of SA_CKPT_WR_ALL_REPLICAS |SA_CKPT_CHECKPOINT_COLLOCATED,
Best regards, Nhat Pham -----Original Message----- From: Girish Nagaraj [mailto:[email protected]] Sent: Monday, December 21, 2015 11:33 AM To: Nhat Pham <[email protected]>; [email protected] Subject: RE: [users] Question regarding shared memory Hi Nhat, Yes, checkpoint created with SA_CKPT_WR_ACTIVE_REPLICA | SA_CKPT_CHECKPOINT_COLLOCATED. Active replica is in active node. So I believe in case of SA_CKPT_WR_ALL_REPLICAS |SA_CKPT_CHECKPOINT_COLLOCATED, it will read from local replica? Regards, Girish -----Original Message----- From: Nhat Pham [mailto:[email protected]] Sent: Monday, December 21, 2015 9:56 AM To: 'Girish Nagaraj'; [email protected] Subject: RE: [users] Question regarding shared memory Hi, I guess your checkpoint was created with flags: SA_CKPT_WR_ACTIVE_REPLICA | SA_CKPT_CHECKPOINT_COLLOCATED. Is it true? If so, for the question below, it should depend on where the active replica locates (active or standby node). Best regards, Nhat Pham -----Original Message----- From: Girish Nagaraj [mailto:[email protected]] Sent: Friday, December 18, 2015 3:25 PM To: [email protected] Subject: Re: [users] Question regarding shared memory Hi, One more question, We have two controllers (active-standby), with the steps mentioned below, when application on the standby node makes saCkptCheckpointRead() call from which replica is data read from? From active or standby replica? We have async collocated checkpoint. Regards, Girish *From:* Girish Nagaraj [mailto:[email protected]] *Sent:* Friday, December 18, 2015 8:38 AM *To:* '[email protected]' *Subject:* Question regarding shared memory Hi, In our checkpoint application we have huge amount of data, some sections have upto 500MB of data, while testing bulk-sync scenario I have observed that performance is very poor. Steps (active-standby model): 1. Start active node and checkpoint 500MB of data 2. Start standby node In 2nd step data is synced from active to standby node during saCkptCheckpointOpen() call, after this application reads data from checkpoint database(10MB at a time) using saCkptCheckpointRead() function call and deposits data to application so that application on standby node is in same state as active. We have observed that saCkptCheckpointRead() call takes lot of time(increases exponentialy? If the total section size is >= 400MB) Any reasoning/performance test results available for checkpoint read/write call? How is checkpoint data transferred between osafckptnd process and application? Is it that agent and osafckptnd share a common shared memory? Regards, Girish -- . -------------------------------------------------------------------------- -- -- _______________________________________________ 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
