On 01/02/2014 06:22 AM, Christine Caulfield wrote:
On 12/28/2013 09:25 AM, Matan Shukry wrote:
As part of my attempt to develop an HA application, I was looking for a
way to ease the development of HA, and I found corosync, which supports
4 C APIs.
Then, I found out corosync is not just an API, but rather an entire HA
system, which I can configure to make my application HA (right?).
However when I attempted to use corosync, I encountered problems:
1. I couldn't find any forum/IRC channel where I can get support. This
mailing list doesn't seem like the best place for this type of question,
I just couldn't find any other place.
2. There was no documentation/example/ on how to use corosync with an
application. Rather, there were _many_ tutorials, each one with
different ways/commands and even different executables, on how to set up
corosync with pacemaker(or heartbeat).
And even though there were many tutorials, they all seemed to show how
to set up an HA nodes; That is, settings up a shared ip address and
shared data. However none showed how to create an HA application.
3. Even after looking, installing, downloading, for over 3 days, I can't
seem to find any documentation on how to use the corosync APIs.
Wikipedia mentioned they exists, and I remember some pdf even showing
sample code on how to use it. But no place showing on how to link
against it, how to install, where is the dll(or should I static link
against the code), and if so, what are the commands syntax/structs. A
lot of missing information which can't simply be guessed.
Hiya,
If you really want to use the corosync APIs to make a cluster-aware
application then this mailing-list is probably as good as place as any
to ask, there's also the #linux-cluster IRC channel on freenode. All
of the APIs are documented in man pages and there are quite a few
examples in the source tree that might be helpful.
However, I'd first check exactly what you want to do, it could easily
be that using pacemaker might be a much easier and more efficient way
of HA clustering your application.
One of the reason that there is little user-level documentation for
the corosync APIs is that they are seldom needed by applications
themselves, the most common ways of clustering things are by the use
of resource and application managers such as pacemaker.
That's not to say you shouldn't use the APIs, depending on what you
need to do, CPG is probably the most likely to be useful if you do
need to do low-level cluster coding.
Just to be a bit more clear, the C APIs (sam and CPG specifically) are
really designed to take applications beyond five 9's of availability,
which is likely the practical limit of an HA solution like Pacemaker.
Unfortunately it has appeared most people don't really care for more
availability then what Pacemaker offers, so there are not many examples
of how to use the CPG applications for application HA.
Regards
-steve
Chrissie
Chrissie
_______________________________________________
discuss mailing list
discuss@xxxxxxxxxxxx
http://lists.corosync.org/mailman/listinfo/discuss
_______________________________________________
discuss mailing list
discuss@xxxxxxxxxxxx
http://lists.corosync.org/mailman/listinfo/discuss
