[hobbit] server load issue, clientupdate bug?

[David Gore]

This consistently causes me to have to restart the hobbit server before 
random false alerts start firing off pages when doing a clientupdate. 
It seems more related to the use of classes in bb-hosts and 
client-local.cfg.  The server is Fedora Core 5.  Clientupdates have 
caused this strange behavior on Itanium 64 HP-UX, OSF DG-UX 4, and 
Solaris 10 client hosts.

~David

David Gore wrote:
> This appears to happen when you try to update files like runclient.sh, 
> hobbitclient.sh, and hobbitclient-`uname -s`.sh or any other file that 
> is not write-able.
> 
> It can or was resolved by making the files write-able, a hassle for a 
> bunch of hosts, and restarting the hobbit server.
> 
> 
> ~David
> 
> David Gore wrote:
>> Henrik,
>>
>> For you consideration, I think the server may go a little crazy if you 
>> try to send updated client packages to too many hosts at the same 
>> time, false failed tests, pages sent on those same false failed tests, 
>> and status pages not available for those same false failed tests:
>>
>> bb-display.log:
>> ... 2006-11-30 01:11:44 Whoops ! bb failed to send message - timeout
>> 2006-11-30 01:11:44 hobbitd status-board not available
>>
>> clientdata.log:
>> 2006-11-30 00:51:20 Whoops ! bb failed to send message - timeout
>> 2006-11-30 00:56:12 Whoops ! bb failed to send message - timeout
>> 2006-11-30 00:58:17 Whoops ! bb failed to send message - timeout
>> 2006-11-30 00:59:41 Whoops ! bb failed to send message - timeout
>>
>> hobbitclient.log:
>> [hobbit at hobbit1 logs]$ cat hobbitclient.log
>> 2006-11-30 00:58:16 Whoops ! bb failed to send message - timeout
>>
>> Here is what my config looks like client-local.cfg:
>>
>> [temip-be-hpux11] # these are class names, not host names
>> clientversion:temip-be-hpux11v10
>> log:/var/adm/syslog/syslog.log:10240
>> [temip-fe-hpux11]
>> clientversion:temip-fe-hpux11v10
>> log:/var/adm/syslog/syslog.log:10240
>> [temip-tns-hpux11]
>> clientversion:temip-tns-hpux11v2
>> log:/var/adm/syslog/syslog.log:10240
>>
>> I suppose I updated about 60+ remote hosts at once, perhaps we should 
>> just try to figure out what our server can handle?  Or should the 
>> server be made to be smarter?  The server is a simple 3 Ghz dual core 
>> Intel Fedora Core 5 host with 1G of memory.
>>
>>
> 
> 
> To unsubscribe from the hobbit list, send an e-mail to
> hobbit-unsubscribe at hswn.dk
> 
>