pica1dilly at yahoo
Jun 19, 2008, 8:50 AM
Post #5 of 7
(465 views)
Permalink
|
|
Re: What best name for the Heartbeat v2 log daemon ?
[In reply to]
|
|
--- On Thu, 6/19/08, Dejan Muhamedagic <dejanmm[at]fastmail.fm> wrote:
> From: Dejan Muhamedagic <dejanmm[at]fastmail.fm>
>
> Aren't both "HA logging daemon" and "Heartbeat logging deamon"
> different enough from "system logging daemon" or syslog?
As a "logging daemon", I would expect that component to replace syslogd.
This is not the case, and that's why it's confusing.
> Linux-HA and Heartbeat have been used for the same thing,
> i.e. a linux HA software.
In my mind, anybody who brings up "Linux-HA" or "HA" to talk about
a cluster manager means "Heartbeat".
There is no confusion with Xen, or OSCAR, or Rocks.
> Of course, I don't mind unifying the name, but
> don't see this as urgently needed.
> There are quite a few more interesting tasks
> regarding the documentation ;-)
I'll explain the urgency by saying that the mismatch between the
documentation and the code, to which you can add alternative options
(i.e. use_logd=yes/no) to achieve the same thing, and recommendations like
"use use_logd=yes" with no further explanations, just prevents from
understanding the true **design** and *intent** of the developers.
Having read the documentation and the source code, I, as developer of
resource agents, have no trust in how ha_log() works today and will keep on
working tomorrow.
Eventually I end up providing one option with ha_log() "to conform to the
standard", regardless of the result returned, and one option without ha_log(),
that guarantees the result I want to achieve while improving the reliability
and the predictability over time
>> Does it make sense to call the HeartbeatLogDaemon, a
>> "log daemon", when it does NOT itself write messages
>> to disk ?
>
> Why not? A syslogd doesn't have to write messages to
> disk either.
>
Ok, I'll take the "syslog" characteristic, describing the
"syslog message protocol", as distinguishing factor
>> Isn't the HeartbeatLogDaemon closer to being
>> itself a "logger" that sends itself messages
>> to the real "log daemon" ?
The fact is though, the component has two parts: ha_logger and ha_logd.
If the daemon lacks the disk-writing, it's missing then one of the main
functions that characterizes as a "logging" daemon.
Which suggests the "ha_logd" component is closer to being an extension of
the "logger" than of a "logging daemon".
> > Then, is the HeartbeatLogDaemon really a "broker" ?
Not a broker, it doesn't negotiate.
> > Isn't it rather a "proxy", or a
Not a proxy, nodes are pushing data to the server,
not that the server is pulling data.
>> "relay" ?
"logger relay" "logger-relay-daemon" sound ok
>> Finally, shouldn't it's temporary FIFO buffer
>> storage feature be evident in the name ?
A relay would appear to know how to store and forward,
rather than drop messages like a saturated gateway or bridge.
Taking your remark above about syslog in account, we get:
"syslog-relay" and "syslog-relay daemon" which sound also fine to me.
What shall it be?
- The Linux-HA "syslog-relay daemon" with the ha_logger ?
- the Heartbeat "syslog-relay daemon" " " " ?
- the Linux-HA "logger-relay daemon" with the ha_logger ?
- the Heartbeat "logger-relay daemon" " " " ?
_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev[at]lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/
|
