[pca] read uname/showrev/pkginfo data from stdin

[Rajiv Gunja]

Asif,
I have run pca against 1200 servers' config files (uname,showrev and
pkginfo) via a for loop to create a list of patches in html format and it
took about 30 to 45 minutes in total. I run a cronjob every 20 minutes to
create a Patch Bundle on our pca-proxy server. When an Engineer submits a
servers' config tar file, it takes about a 75 to 120 seconds depending on
how old the server is and when it was patched last. I have chosen a 20
minute interval as a maximum of 12 servers can be processed within that
time. On an average we have about 5 servers submitted for patch bundle
creation.

Of course the time taken by your wrapper script to generate a bundle from
pca's output will depend on how far back your server is from the current
patchdiag.xref cutoff date.

For example I created Patch Bundles for 4 Solaris 10 servers this morning
and my cronjob took about 11 - 12 minutes for the all 4 of them as each of
the bundle created was about 1.2 GB. They were all Sol 10 U3 and had never
been patched since they were released into production.

PCA is very fast compared to any product out there SUN's or 3rd party and
because we can write wrappers around it, PCA becomes the most valuable tool
for Solaris administrators.

We used to use TLP until end of 2007 and it would take us anywhere from 4
hours to 2 days for our patch bundles to be created depending on the load on
the TLP servers. Camparitively, on our PCA-proxy a Patch bundle usually
takes about 1 to 3 minutes.

-GGR

--
Rajiv G Gunja
Blog: http://ossrocks.blogspot.com


2009/5/11 Asif Iqbal <vadud3 at gmail.com>

>
>
> On Mon, May 11, 2009 at 3:37 AM, Martin Paul <martin at par.univie.ac.at>wrote:
>
>> Asif Iqbal wrote:
>>
>>> Is there a way for pca to read the uname/showrev/pkginfo  data from stdin
>>> instead of from a file?
>>>
>>
>> No, there isn't.
>>
>>  I can make 500 /tmp/hostname dir and have the pca read it from each of
>>> those
>>> /tmp/hostname files.
>>>
>>> But it would be lot faster if I could feed that data directly to pca
>>> through
>>> stdin.
>>>
>>
>> I'm not sure whether this would really speed things up. Creating the files
>> in a tmpfs and reading from them should not be a lot of overhead, compared
>> to the time pca itself needs for the analysis.
>>
>
> How about running multiple copies of pca since I will have about 500
> /tmp/hostnameX  dirs to analyze?
>
>
>
>>
>> Martin.
>>
>>
>
>
> --
> Asif Iqbal
> PGP Key: 0xE62693C5 KeyServer: pgp.mit.edu
> A: Because it messes up the order in which people normally read text.
> Q: Why is top-posting such a bad thing?
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.univie.ac.at/mailman/private/pca/attachments/20090511/5ac205ce/attachment.html