[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[Linux-cluster] Linux Cluster Questions



Hello everyone,

I'm new to this list so please bear with me.  My previous Linux
clustering/HA has come from LinuxHA and Linux LVS.  Linux HA/LVS are great
for application HA/Load balancing but not for HA storage, before you all
scream I know about DRBD but that seems slow and kludgey.  I'm wanting to
setup two servers with CLVS and I want to create an LVM2 mirror spread
across the two servers (neither server will access the partition at the same
time -- so I'm not too worried about using EXT2/3).  Ideally I want to have
a 'global' Volume Group with PVs from both boxes and from that create the
mirrored LVs.  

If this is crazy or undo-able please let me know.  Heh, I'm not sure if I'm
making things worse for myself or not but I'm running on Debian 4.0 Etch.

>From what I've gathered I need to configure the following daemon and
applications, let me know if I'm missing anything (or have too much =):
1. CCS - Cluster configuration manager
2. CMAN & DLM -- Cluster Manager (or GULM?)
3. Fence I/O
4. CLVM daemon 
5. gnbd -- maybe for fencing purposes?

Some questions I have about Linux Cluster for CLVM are:
1. lock_gulmd vs. cman & dlm  -- If I were using GFS2 I would want to use
glumd right? 

2. fencing -- Ick!!! I don't have any of the fencing agents -- such as an
APC MasterSwitch, IBM Blade center, Brocade McData, Sanbox/Qlogic, or Vixel
fibre switches, Egenera blades, HP ILO ... that leaves me with GNBD and
Manual/Manual_ack.  From the man pages it's suggested I not use
Manual/Manual_ack for production.  I really don't need automatic re-adding
of an node once it comes back up, what issues will I run into using manual
fencing?

3. cman -- What issues are there to getting CMAN running?  I'm having
trouble getting a custom kernel compiled with the CMAN-Kernel module.

4. ccs_tool, cman_tool and version numbers.  I see there's config versions
in /etc/cluster/cluster.conf and that ccs_tool is used to send updates
(changes) to the other nodes.  Soooo I assume when you make a change to a
running cluster you increment the version number each time and you use that
new version number when you run cman_tool?


Thanks,
Jim







James Miller - MCSE RHCE CISSP
Sr Systems & Network Administrator
Simutronics Corp.
www.play.net
636.946.4263 x113 


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]