Rajesh G wrote:
> Hi Vassun,
>
> It was really nice to see the description of the patchadd behavior. 
> I have a question. What will be the performance impact of using "patchadd -t" 
> to "patchadd", if I try to install one patch at a time. This is required for 
> me, as I want to update the system and my code is dependent on the old 
> patchadd return codes. Can you please let me know.
>
> Thanks,
> Rajesh
>  
>  
> This message posted from opensolaris.org
> _______________________________________________
> install-discuss mailing list
> install-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/install-discuss
>   
Hi
I have never done a comparison ( might do next week )
it kinda depends on number of zones and number of patches.
for patchadd -M we mount and unmount all zones twice, irrespective of 
how many patches are being added.

for patchadd -t in a for loop, then we mount/unmount twice for every patch.
Above is assuming some or all zones are halted. Note that zone mount and 
unmount is not a major operation, so if only addign say 5 patches to a 
system with 2 zones, then there will be little difference anyway.
if you have a large number of zones, then it will a long time anyway ( 
just longer with the -t ) either way.

Note When we patch halted zones with a patchadd patch higher than 
119254/119255 rev 14 ( suggest taking hihest rev available at all times 
) then zones are not actually booted, rather they are "mounted", this is 
a faster operation overall than booting.


I would suggest adding latest rev of 119254/119255 first
run patchadd -a -M on the directory of patches. The -a does a dryrun ( 
does not modify system )and will mount any zones and do any dependency 
checks, so it might catch certain failures upfront. Note that some 
patches are not compatible with -a ( 
120011-14/120012-14/118833-36/118855-36)
Then run the patchadd -t assumign the patchadd -a -M all worked ok.
Enda

Reply via email to