HyperFabric Release Notes (B.11.11.04.02)

HyperFabric Release Notes

Patches and Fixes in This Version

Chapter 1 15

CR JAGaf21296 1.Retransmission failures are now marked as FAILED in the SKGXP

request handle and put it in the done queue. We are not returning

SKGXP_FAIL to Oracle which would cause Oracle crash otherwise

during node crash/TOC/CLIC daemon core dump.

2.VCs were not being failed after a node crash/TOC/ CLIC daemon core

dump. This will result in peer Oracle no being able to establish CLIC

connectivity. Clic shutdown on both the nodes only was the alternative.

This is ﬁxed. Fix for VC/Node crash issue needed to initiate VC

establishment information to be always sent only after CONN RSP is

done.

3.Compatibility ﬁxes for node crash and few logging.

4.Check for node crash which will never come up ﬁx to fail VCs via

Software heartbeat -or-- via encoding in the connectivity message.

CR JAGaf21311 CLIC_VC_FREEZE ioctl was modiﬁed. It was not doing its job for

MSGIDU to read. Protocol error handling without setting bulk

accumulator. This was resulting in update sequence/Out of order

resulting in Oracle crash.

CRJAGaf21313 Fix for Incompatible mode of operation appearing wrongly in the

clic_log.

CR JAGaf21317 Using possible failover ﬂag to do failover retry++. Possible failover is

tracked in the APP NQ CB now. Temp storage of EPT and revert it back

in the ‘clic_csi_choose_vc’ path. Failover ﬂag and retransmission states

are converted into volatile storage speciﬁer to not have any memory

optimizations.

CR JAGaf21319 Software HB redundant factors and their min default are downsized to

fail within 45 Seconds.

CR JAGaf21320 MSGIDU_RSP_DONE path failing VCs and setting failover ﬂag order

is changed. This could result in Oracle crash.

CR JAGaf21322 Fix for the bug in handling RESUMED active element ﬂag. After

multiple failed and resume events Oracle could crash because of this

issue.

Table 1-3 Fixes Included in this Release (Continued)

CR Number Description