Apply Bundle patch in HPUX ServiceGuard issues

[mark yeo]

hi, i am applying HPUX Bundle Patches to my 3 HPUX machines,

dev - HP-UX uxipos03 B.11.00 U 9000/800 1503746527 unlimited-user license
qas - HP-UX uxipos02 B.11.00 U 9000/800 1507746547 unlimited-user license
prd - HP-UX uxipos01 B.11.00 U 9000/800 1533746527 unlimited-user license

Below is the steps which i am planning ahead for DEV.

1. Verify that the bundle has downloaded correctly to your system by using swlist. For example, to verify the GOLDQPK11i bundle, enter:
swlist -d @ /tmp/GOLDQPK11i_11.11.depot

2. As root, use swinstall to install the patch bundle and specify the patch_match_target option (for HP-UX 10.20, use match_target):
swinstall -s /tmp/GOLDQPK11i_11.11.depot \
-x autoreboot=true -x patch_match_target=true

if the system is able to boot up and i am able to start both of my oracle and SAP, then everything is a success.

Before applying the patches, i will do the below first.

3) Full file backup command
fbackup -f /dev/rmt/c0t3d0BEST -i / -e /usr/sap/trans

4) To do make recovery tape:
# make_tape_recovery -x inc_entire=vg00 -v -a /dev/rmt/0mn
4.1) is the above make_tape_recovery command good? or should i use "Make_tape_recovery -I -v -a /dev/rmt/0mn -x inc_entire=vg00"
4.2) is the choice of device an issue here? 0mn as instructed or should i use the one from my fbackup which is c0t3d0Best? see (6) for my list of devices in /dev/rmt/


5) To do full system backup:
# fbackup -vf /dev/rmt/0m -i /
5.1) i dont think that this will work as the whole backup of the filesystem is going to be greater than the disk space of the tape itself, hence i have to exclude my filesystem as evidenced in (3)

6) list of devices
# cd /dev/rmt
# ls
0m 0mn c0t3d0BEST c0t3d0BESTn c0t3d0DDS c0t3d0DDSn stape_config
0mb 0mnb c0t3d0BESTb c0t3d0BESTnb c0t3d0DDSb c0t3d0DDSnb

Finally, my QAS and PRD are on a one-package ServiceGuard. Will the above steps for DEV works for them as well?

Thanks

[Massimo Bianchi]

Hi,
some little notes:

2. As root, use swinstall to install the patch bundle and specify the patch_match_target option (for HP-UX 10.20, use match_target):
swinstall -s /tmp/GOLDQPK11i_11.11.depot \
-x autoreboot=true -x patch_match_target=true "*"

( you forgot to specify the product!)


If the system boots up is a nice thing, but it's not sufficient:


check that all fileset are configured correctly:


swlist -l fileset -a state | more


if there is something that is NOT in state "configured", check...

Check also all the functionality, not only the startup: for example, once i had a patch that modified the remsh behavious, leadoing to some sessions remaining appended. Check all the shell script, and wait at least a week before going to production.

you should use sintax like in 4.1 for the make_tape_recovery. Omn or C3.. is exactly the same...



for point 5: I agree, but remember to backup the trans separately.

And finally: my QAS and PRD are on a one-package ServiceGuard. Will the above steps for DEV works for them as well?


YES: but it' best if you shut down the cluster before applying the patchset, since normally sw issue a reboot, rather than a shutdown -ry. SAP/ORacle do not likes it much ;)


HTH,
Massimo

[melvyn burnard]

whatever you do, issue the cmhaltnode on the node BEFORE installing the patches, as if they require a reboot, it issues a kill -9, and you will see the node TOC rather than reboot.

[RAC_1]

Apart from stopping the node, I would also consider running swinstall in preiview mode before actually installing the patches.

Anil

[mark yeo]

A) Are the below two equivalent?
A.1) swinstall -s /tmp/GOLDQPK11i_11.11.depot \
-x autoreboot=true -x patch_match_target=true "*"

A.2) swinstall -s /tmp/GOLDQPK11i_11.11.depot \
-x autoreboot=true -x patch_match_target=true

A.3) the difference is with the "*" at the last entry.

B) If after the patches, i cant boot up then i had to use the make_tape_recovery as belows

B.1) get into PDC, press ESC,
B.2) boot hardware path (i can get this from ioscan -fn | grep -i tape"
B.3) and the system hopefully can be rebooted and is fine thereaftere.

B.4) for make_tape_recovery, which one should we use
B.4.1) make_tape_recovery -x inc_entire=vg00 -v -a /dev/rmt/0mn
B.4.2) Make_tape_recovery -I -v -a /dev/rmt/0mn -x inc_entire=vg00

C) When then will we need to do a recover from fbackup?

C.1) Can we use the following frecover command to recover from a fbackup?
C.1.1)fbackup command is fbackup -f /dev/rmt/c0t3d0BEST -i / -e /usr/sap/trans

C.1.2)frecover command is frecover -f /dev/rmt/0m -rv

Thanks, guys...