gtps2m31ACF/SNA Data Communications Reference

Hardware IPL Considerations

If HARDREC=YES is coded in SNAKEY on restart, TPF attempts to resynchronize the VR (virtual route) sequence number for each channel-attached NCP. This is due to the VR sequence number not being keypointed before the hardware IPL. The NCP must be version 4 release 2 or above to support VR resynchronization.

When VR resynchronization is in progress, TPF continues to receive inbound PIUs and hold all outbound PIUs on VR0 in core blocks until VR Resync is completed. Meanwhile, TPF sends a few SNA commands on VR1 in an attempt to recover the next VR sequence number from the NCP. It is critical that the amount of inbound traffic should be minimized, because TPF must queue all outbound PIUs on VR0 in core blocks. The VR PIU Pool threshold in the gateway NCP is used to control the network traffic between TPF and NCP.

To alleviate TPF resource problems, TPF performs VR resynchronization serialization across all channel-attached NCPs. The SNAKEY parameter VRRTO is used to specify the time-out value for VR resynchronization for each NCP. Upon timeout, TPF either declares that VR resynchronization is complete for the NCP if the response to DACTVR is outstanding, or TPF shuts down the NCP because of VR resynchronization failure.

The VR PIU Pool threshold should be set properly to prevent the NCP from depleting its buffer pool and to minimize TPF congestion from queuing core blocks.

The NCP generation parameter, VRPOOL, is the parameter for the VR PIU pool. See NCP/SSP/EP Resource Definition Reference for more information about VR PIU Pool:

Virtual route resynchronization does not address the problem of LU-LU session synchronization following a soft or hard IPL. For NEF type terminals connecting using NEF V2 or ALCI, no session resynchronization is necessary. This is because NEF does not check session sequence numbers, nor does it use brackets or change direction. For X.25 terminals, NPSI does not use brackets or change direction, but it does check session sequence numbers. Session resynchronization (CLEAR/SDT) is required to recover the NPSI LU_1 sessions. For SNA-type terminals, because the session control block (RVT2) is not keypointed on a hard IPL, the LU session may be hung; that is, TPF status does not agree with LU status. TPF is aware that a session may exist. (LU control block, RVT1, is keypointed on a time-initiated basis, and LU information is preserved.) However, TPF is not aware of the status of the session; that is, in brackets, response pending, or next expected sequence number.

TPF does provide time-out utilities to detect hung sessions and attempt resynchronization based on LU type (CLEAR/SDT or CLEAR/STSN/SDT) which should recover most, though not all sessions; for example, LU6.2 sessions receiving unexpected sequence numbers are terminated.

If HARDREC=NO is coded in SNAKEY on restart, TPF issues a DISCONTACT request for each NCP using FID4, if the NCP shows active in the RVT entry. As a result of DISCONTACT, automatic network shutdown (ANS) takes place. The network has to be restarted. The DISCONTACT is issued because TPF cannot determine the correct VR sequence numbers, as the control block containing the sequence numbers (in the subarea address table) was not keypointed prior to the soft or hard IPL.