Re: [webmin-l] webmin cluster updates/hanging rpc.cgi

[Pat E. <mai...@pa...>]

hi,

On Tue, 03 Oct 2006 20:03:10 +0200, Jamie Cameron <jca...@we...>  
wrote:
>> while celebrating my 365th day of uptime ;) i found a lot of
>> hanging/waiting rpc.cgi processes from webmin versions removed long,  
>> long
>> time ago (ie:
>>
>> # ps ax|grep rpc
>> 18103 ?        S      0:00 /usr/local/webmin-1.270/rpc.cgi
>>
>> when 1.300 is installed. maybe you should kill them when upgrading.
>
> That is odd - the upgrade does indeed kill all processes for the old
> version. Do you see any from versions newew than 1.270?
yes, i saw, processes from each version up to 1.300 - and they were a lot,  
around 10-20 per version i think (but i killed them meanwhile)

>> which reminds me, it is quite hard to cluster-update webmin, because you
>> limit this functionality to servers which are interconnected with the
>> "fast rpc mode" which seems to be in the end nothing more, as that i  
>> need
>> to keep up some ports above 10000. now, i'm in the situation, that i
>> cannot open more than some 20 ports in the hardware firewall my provider
>> offers me and stupidely - i canot open ranges, so i cannot connect my
>> webmin servers "fast" in the webmin sense although the speed between  
>> them
>> is 100mbit. couldn't you make this limitation optional?
>
> Sadly no .. only in fast mode can there be a persistent connection from
> the master to the slave, which is needed for the upgrade.
ok, as i see you send the update package from the master to the slave -  
couldn't you create another update mode in which each clustered server  
would get the signal "update now" as i do it now by hand?

PAT