On Sun, 17 Mar 2024, Petr Vokac wrote:
> On 3/15/24 16:42, Marcus Ebert wrote:
>> One correction:
>>
>> On Fri, 15 Mar 2024, Marcus Ebert wrote:
>>
>>> Setup is
>>> - redirector rdc-redirector.belle.uvic.ca
>>> - 3 servers xrd{1..3}.belle.uvic.ca
>>>
>> servers are xrd{4..6}.belle.uvic.ca
>
> Simple macaroon token request takes 30s?! You should probably start with this
> issue ... actually this is pretty simple, your IPv6 connectivity is broken
> and this could bring with a lot of fun especially with old HTTP libraries
> like libneon that are still in by KEK FTS server (actually for better IPv6
> support it would be necessary to enable `curl` for FTS/gfal and run FTS on
> EL8 or EL9 and that OS version is not yet officially supported).
>
> You should really fix HTTP-TPC pull mode, which should be preferred (btw: FTS
> developers now considering changes in FTS that will prevent automatic
> fallback push <-> pull in case one method doesn't work, because this
> functionality just hides serious site issues and makes HTTP-TPC transfers
> debugging unnecessary complex)
>
Right. Redirector and servers are in different subnets and there have been
issues with the routing using IPv6. For now we disabled all IPv6 and
removed the DNS entry for it, focusing only on IPv4 to see if that works.
Later on we can enable IPv6 again once all routing issues are fixed.
Indeed, removing IPv6 and the DNS entry for it makes pull work now. At
least in most cases. Only issue that remains is with a KEK server while
all other sites can transfer from that KEK server just fine.
The pull still doesn't work with that server and for push FTS gets a "
SocketException while pushing.... Connection reset by peer (Write
failed)". It doesn't happen for all transfers but for
most.
An example fts log for this kind of error can be found at
https://particle.phys.uvic.ca/~mebert/xrootd/fts-kek.txt
and the parts of the local log are in
https://particle.phys.uvic.ca/~mebert/xrootd/redirector.log
https://particle.phys.uvic.ca/~mebert/xrootd/xrd6.log
https://particle.phys.uvic.ca/~mebert/xrootd/xrd5.log
(in case things are missing in the logs, please let me know how I can get
the output corresponding to a single request)
Cheers,
Marcus
########################################################################
Use REPLY-ALL to reply to list
To unsubscribe from the XROOTD-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
