Mailing List Archive: Data service monitoring

Hello all!!
I'm introduce my self like juampe.
I Ask, first to my self and second to the list.
There are support for Data service monitoring?
If aren't, well I believe that heartbeat test the machines but not the
data services (please tell to me if I'm in a mistake).
I mean that the Data services (ex: Apache) is the essence of availablility.
If the machine is OK an the data service is crashed? Bad thing...
I think that is eassy to test the data service (Like HA-NFS in SUN).
Simply send request from the nodes in the cluster to the data service and
wait a familiar response

isn't enought? what do you think?

for the apache
apache "GET / HTTP/1.0" "HTTP/1.1 200 OK*"

The first string is a data service to be monitored
The second string is a query to the data service
The third is a familiar response or a pattern

In other cases I believe will be more complex...

Saludos
--
Debian GNU/Linux 2.1 Kernel 2.2.14 Pentium 200 64Mb/4Gb

_ | There are no need to worry about it
/ \ | TuX Knows what happens...
| () | | Linux User Nº 89286 -- GUPLE Member -- ICQ 25811047
| \_/ | juampe-ARROBA-retemail-PUNTO-es
\ | http://webepcc.unex.es/~juampe
\ | Yo también 3000ya.com http://www.3000ya.com
debian | TARIFA PLANA ¡YA!

Hi Juan,

Juan Pedro Paredes wrote:
>
> Hello all!!
> I'm introduce my self like juampe.
> I Ask, first to my self and second to the list.
> There are support for Data service monitoring?
> If aren't, well I believe that heartbeat test the machines but not the
> data services (please tell to me if I'm in a mistake).
> I mean that the Data services (ex: Apache) is the essence of availablility.
> If the machine is OK an the data service is crashed? Bad thing...
> I think that is eassy to test the data service (Like HA-NFS in SUN).
> Simply send request from the nodes in the cluster to the data service and
> wait a familiar response
>
> isn't enought? what do you think?
>
> for the apache
> apache "GET / HTTP/1.0" "HTTP/1.1 200 OK*"
>
> The first string is a data service to be monitored
> The second string is a query to the data service
> The third is a familiar response or a pattern
>
> In other cases I believe will be more complex...
>
> Saludos

This is a piece of the puzzle for the section in the to do list called
"Diagnostic Activities". There is a whole strategy for how to do diagnostics
which hasn't been worked out yet, but it still needs to be.

There are at least two components needed:
1) probes, like the ones you discussed
2) decision making software to decide what to do
with the probe results

-- Alan Robertson
alanr@bell-labs.com