pcp
[Top] [All Lists]

RE: [pcp] JSON PMDA

To: "'Frank Ch. Eigler'" <fche@xxxxxxxxxx>
Subject: RE: [pcp] JSON PMDA
From: "Ken McDonell" <kenj@xxxxxxxxxxxxxxxx>
Date: Fri, 17 Apr 2015 13:16:39 +1000
Cc: "'David Smith'" <dsmith@xxxxxxxxxx>, <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <20150416233508.GF6822@xxxxxxxxxx>
References: <448002717.7934024.1427683964254.JavaMail.zimbra@xxxxxxxxxx> <552699FE.7040801@xxxxxxxxxx> <2139482617.15593599.1428634701360.JavaMail.zimbra@xxxxxxxxxx> <552D6524.1030803@xxxxxxxxxx> <y0mfv80ubzj.fsf@xxxxxxxx> <552FD368.6030602@xxxxxxxxxx> <20150416152834.GB6822@xxxxxxxxxx> <003f01d07887$90ee5180$b2caf480$@internode.on.net> <20150416222728.GD6822@xxxxxxxxxx> <005301d0789a$2ecae550$8c60aff0$@internode.on.net> <20150416233508.GF6822@xxxxxxxxxx>
Thread-index: AQEsFordHoUvuMtbVEHoa9Ri4iN7VgII2YlEASYuiQEBTX5F1AIBL5NAAVeb5C0CxhiccgJtj29oAla44F4B/NxIjQKEmVZunfpCVEA=
G'day Frank.

> (AIUI the status quo is that such a pmda restart involves a pmcd restart,
so
> then pmie, pmlogger, and the other pmcd-connecting tools you listed would
> notice and stop or pmLookup* upon reconnect.)

As Nathan noted earlier, the status quo is (IMHO) broken ... in the
classical PCP world, pmcd kept running as PMDA's were installed, removed,
died or restarted.  It is only relatively recently that the "don't run pmcd
as root" mantra caused changes in pmcd behaviour so that pmcd is now more
likely to be stopped and restarted.

And this would indeed cause pmie down that path.

But my test was
- start pmie fetching simple.now
- kill the simple PMDA
- send a SIGHUP to pmcd, which restarts the simple PMDA

In this scenario pmie does not go through the reconnect path.

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