[OpenIndiana-discuss] pkg info slowness/ram usage

[Nikola M]

On 03/14/16 08:25 AM, Bryan Wann wrote:
> Hi all,
>
> I’m preparing a new OI hipster-2015 machine to replace my old OI 151a2 box. I am running Chef on both with recipes to keep a couple dozen IPS package up-to-date. I noticed checking packages was very slow, dug into it and noticed the underlying “pkg info” command is very slow to return, 7-8 seconds. Watching prstat I also noticed the pkg process grows very quickly to 400-500 MB RSS before it returns. This seems like a lot of RAM to be eating up for a simple info command (which gets repeated each time as Chef checks each package).
>
> Compared to my old 151a2 box, the same command there returns in ~1 second and only consumes around 40 MB of memory. I’m not using any exotic pkg publishers, just the standard openindiana.org/hipster-2015 and sfe repos.
>
> Any ideas of what may be causing this slowness and/or memory usage?

Could be that hipster acccumulates many frequent changes over time, 
since it is always updating something e.g. rolling-release and that gets 
up RAM usage up for pkg tasks.
This issue is solved with hipster releasing ISO and re-publishing and 
changing repository (to,say hipster-2016) and planned landing into /dev 
.  It would be great if publishing into /dev could coincide with Hipster 
move to new repo name.

IPS is also updated by Oracle and it is assuming there is at least 1GB 
of RAM for IPS pkg operations. One can dig into IPS source and see if 
something can be done to improve it.

If you have recepies to keep IPS packages up to date,
you can contribute them into OI hipster and have them built and tested 
by wider audience and re-built for releases etc,
so you don't have to do it yourself and packages are always there.
There is also 'pkgbuild' project that is building SFE packages for OI 
for hipster (localhostoih, http://sfe.opencsw.org/localhostoih/) so that 
is also place for updating packages contribution. (as well as 
hipster-encumbered, http://pkg.openindiana.org/hipster-encumbered/).