pcp
[Top] [All Lists]

Re: [pcp] pmcd restart race condition

To: Nathan Scott <nathans@xxxxxxxxxx>, Martins Innus <minnus@xxxxxxxxxxx>
Subject: Re: [pcp] pmcd restart race condition
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Tue, 6 Oct 2015 16:11:02 +1100
Cc: PCP <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <2080298045.49282349.1444088161969.JavaMail.zimbra@xxxxxxxxxx>
References: <560EB93A.6020606@xxxxxxxxxxx> <y0my4flump3.fsf@xxxxxxxx> <00bd01d0fd56$189950a0$49cbf1e0$@internode.on.net> <5612D405.2030701@xxxxxxxxxxx> <5612DAA5.9090403@xxxxxxxxxxxxxxxx> <2080298045.49282349.1444088161969.JavaMail.zimbra@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.3.0
On 06/10/15 10:36, Nathan Scott wrote:
...
The problem was reported by someone who wanted to be able to spin up VMs
quickly, which were preconfigured to have pmcd installed and a (at least
one) custom PMDA which had a slow Install.  For some reason they couldn't
make the Install faster, it was impractical for them to pre-configure the
PMDA (namespace install, etc) because of the way they prepared the system
images & Install was measured as being the longest part of their instance
bootup (10 - 20 seconds IIRC).

Since we have in-tree PMDAs with long check_delay ./Install settings too,
this seemed like something we should be doing in the background anyway.

I would question if PCP is important, why is so important to get pmcd running with only some of the relevant PMDAs installed, however ...

I've re-audited the rc script, and with some minor re-ordering of tasks near the end of the "start" action, I have something that I now believe is safe to run _pmda_setup in the background again.

This passes qa -g pmcd on one machine for me, so I'm going to commit it shortly.

<Prev in Thread] Current Thread [Next in Thread>