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

Re: [HTCondor-users] HTCondor Transferer issue for HAD Replication issue with Shared Port



> On Apr 18, 2018, at 10:35 AM, Oliver Freyermuth via HTCondor-users <htcondor-users@xxxxxxxxxxx> wrote:
> 
> Am 18.04.2018 um 17:29 schrieb Todd L Miller:
>>> Can somebody tell me why the transferer tries to connect to an arbitrary port?
>> 
>>     Either the transferer to which it's connecting isn't configured to use shared port, 
> 
> Our configuration is enforced to be identical between both Central Managers by Puppet, so both have the configuration stanzas included in my last mail. 
> So unless there is something that needs to be done in addition specific to the TRANSFERER, everything should be configured for shared port usage. 
> 
>> or because the transferer was never updated to be able to use shared port.  At first glance, it looks like the transferer creates a new listen socket rather than use its existing (shared port) command socket.
> 
> That would be pretty ugly - I can't believe anybody would want to run Central Managers, which need to be in somewhat public networks since all schedds connect to them, without a firewall... 
> What can I do? 


Starting with HTCondor 8.7.4, the transferer code in HAD works with shared port (and IPv6), and no longer binds its own TCP port.

https://htcondor-wiki.cs.wisc.edu/index.cgi/tktview?tn=6413

Thanks and regards,
Jaime Frey
UW-Madison HTCondor Project