WILLIAMS_B@svax05.pfizer-central-research.co.UK (09/09/87)
Does anyone know how you can relate the first few digits of the PID to the cluster node you are on? I'd like to be able to tell which account file I'm looking at by the PID's of the processes therein. Thanks Brian Williams (address above)
nagy%warner.hepnet@LBL.ARPA (09/10/87)
>Does anyone know how you can relate the first few digits of the PID >to the cluster node you are on? I'd like to be able to tell which >account file I'm looking at by the PID's of the processes therein. The following was pulled from the SYS$LIBRARY:LIB.REQ file (under VMS V4.5): ! ! *** WARNING - THE INTERNAL STRUCTURE OF THE EPID IS SUBJECT TO RADICAL ! *** CHANGE BETWEEN VERSIONS OF VMS. NO ASSUMPTIONS SHOULD ! *** EVER BE MADE ABOUT ITS FORMAT ! PCB$V_EPIC_PROC size=21 bits offset=bit 0 ! PROCESS ID FIELD, CAN CONVERT TO PCB$L_PID PCB$V_EPID_NODE_IDX size=8 bits offset=bit 21 ! IDX - INDEX TO TABLE OF NODE IDENTIFICATIONS PCB$V_EPID_NODE_SEQ size=2 bits offset=bit 29 ! SEQ - SEQUENCE NUMBER FOR NODE TABLE ENTRY REUSE PCB$V_EPID_WILD size=1 bit offset=bit 31 ! FLAG THAT EPID IS WILDCARD CONTEXT FOR $GETJPI, ! AND NOT A VALID EPID So, the node is represented in the EPID (External or Extended PID), by an index into a table. I believe that nodes are added to the table in the order they join the VAXCluster. Thus the node identified by a particular slot in the table can possibly change from boot to boot. My interpretation of this is that the EPID is not useful in determining which processor a process ran on in the past; only which processor it is running on at the present time (present incarnation of the VAXCluster after a boot). = Frank J. Nagy "VAX Guru" = Fermilab Research Division EED/Controls = HEPNET: WARNER::NAGY (43198::NAGY) or FNAL::NAGY (43009::NAGY) = BitNet: NAGY@FNAL = USnail: Fermilab POB 500 MS/220 Batavia, IL 60510