strange scenario setting up MMM via a cloud environment

[Tomas]

Hi experts,

I want to ask you couple of questions of MMM 1.2.6 while I try to set-
up it in a Cloud network.
which provide us special service to use floating private IP through
those servers.

I encounter something which I never meet during test in a local sub
network, I just want to know if they are normal or not.

1. I using libnet 1.1.4 source code, compile it into a 64bit red hat,
which provide me "libnet.so.1" to help compile send_arp ,but I still
need copy the libnet.so.1 to /lib64/libnet.so.1 manually , Is that
right step for 64bit redhat Ent 5?

2. to my understand, the connection between servers in the cloud seems
ok, ping result is always between 10 and 100ms, but MMM have a lot of
error like below:

[2009-10-20 17:03:37]: 4834: Listener: Waiting for connection...
[2009-10-20 17:03:37]: 4834: CHECKER: ping: ERROR
[2009-10-20 17:03:49]: 4834: Listener: Waiting for connection...
[2009-10-20 17:03:52]: 4834: Listener: Waiting for connection...

the thing makes me curious, is the error seems happen always at
checker ping, but check mysql and check rep seems always ok, or
sometimes debug log also show
[2009-10-20 17:03:52]: 4834: Listener: Waiting for connection...

but manually connect always ok, I do not know why :(

and "mmm_control show" always show the status was bouncing between
HARD_OFFLINE and AWAITING_RECOVERY, I am not sure it is because
network of cloud is not stable, or some package support MMM is not
properly installed ? I tried fping and send_arp manually , they seems
both can work ...

3. when I am lucky and the network is good enough for db1 and db2 get
online in "mmm_control show" command, switch a role will spend over 3
minutes, that is too long for a smooth failover/switchover...
IS that because configure issue or network instability? I never meet
this in a sub network environment. role changes always done in 1
minutes.

4. I even have a situation looks like below during a Role switch,
totally let me blue, of course, write ip 10.xx.xx.21 can not
reachable :
db1(10.xx.xx.xx): master/ONLINE. Roles: reader(10.xx.xx.20;), reader
(10.xx.xx.22;)
db2(10.xx.xx.xx): master/HARD_OFFLINE. Roles: writer(10.xx.xx.21;)

5. to my observation, MMM seems too sensitive for a cloud network
environment, states change frequently by network impacts, even from
personal feeling, the cloud network is quite stable. Is that because
cloud network, or something need I pay more attention about the 64bit
red hat 5?


Please give me some hints

Thanks a lot.

Yours Tomas

[Tomas]

I use libproc-daemon-perl_0.03.orig.tar.gz and Algorithm-
Diff-1.1902.tar.gz also.

I want to make sure it is not a issue caused by libnet1.1.4 .

And even the debug information shows monitor can not reach the host of
agent,
manually login , ping, mysql check, mysql rep check always fine.

I need your advice :)

Thanks a lot.

[Tomas]

and which fping version you use ?

should i use the latest one ?