You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently we are testing dCache doors with standard pool redirection for transfers.
I noticed that unlike with DAVs I must make sure the pool’s IPs must be on /etc/hosts at the door to be able to make the transfer work.
Why I do not have to do this for DAVs to work?
See below:
For xrootd doors tests external and internal transfers need the door to be able to resolve to external IP address pair of the pool or the internal depending on the clients network locality the:
Example external client access via xroot
=============
1 Nov 2024 11:03:42 (Xrootd-dcint-door002-externalipv6) [door:Xrootd-dcint-door002-externalipv6@xrootd-dcint-door002Domain:AAYl2z2eJcA] Unable to resolve IP address 2620:0:210:1:0:0:0:80 to a canonical name
=============
The pool's DNS resolution is at the door is:
[root@dcint-door002 ~]# ping dcint-pool04
PING dcint-pool04(dcint-pool04.sdcc.bnl.gov (2620:0:210:8803::112)) 56 data bytes
64 bytes from dcint-pool04.sdcc.bnl.gov (2620:0:210:8803::112): icmp_seq=1 ttl=61 time=0.262 ms
64 bytes from dcint-pool04.sdcc.bnl.gov (2620:0:210:8803::112): icmp_seq=2 ttl=61 time=0.247 ms
[root@dcint-door002 ~]# nslookup dcint-pool04
Server: 10.42.34.6
Address: 10.42.34.6#53
Name: dcint-pool04.sdcc.bnl.gov
Address: 10.42.38.112
Name: dcint-pool04.sdcc.bnl.gov
Address: 2620:0:210:8803::112
Transfer works, if explicitly the pool external pair is enabled to resolve at the door
[root@dcint-door002 ~]# cat /etc/hosts|grep pool04
2620:0:210:1::80 dcint-pool04.sdcc.bnl.gov dcint-pool04
192.12.15.128 dcint-pool04.sdcc.bnl.gov dcint-pool04
Reload nscd
systemctl reload nscd.service
Check the external pair is resolved
ping6 dcint-pool04
ping dcint-pool04
The xroot door needed to be reloaded
[root@dcint-door002 ~]# systemctl list-dependencies dcache.target
dcache.target
● ├─dcache@webdav-dcint-door002_httpsDomain.service
● └─dcache@xrootd-dcint-door002Domain.service
[root@dcint-door002 ~]# systemctl restart dcache.target
[cgamboa@lxplus914 ~]$ gfal-copy -f /etc/services root://dcint-door002.sdcc.bnl.gov:1096/pnfs/usatlas.bnl.gov/cgamboa/test.root.write.cern.secure.1
Copying file:///etc/services [DONE] after 2s
For secure xroot the IPV4/IPV6 counter parts needs to exits in the /etc/hosts resolution otherwise
[cgamboa@spar0101 ~]$ gfal-copy -f /etc/services roots://dcint-door002.sdcc.bnl.gov:1096/pnfs/usatlas.bnl.gov/cgamboa/test.root.write.cern.secure.1
Copying file:///etc/services [FAILED] after 1s
gfal-copy error: 40 (Too many levels of symbolic links) - Error on XrdCl::CopyProcess::Run(): [FATAL] Redirect limit has been reached: (destination)
At the door,
01 Nov 2024 11:35:22 (Xrootd-dcint-door002-internal) [door:Xrootd-dcint-door002-internal@xrootd-dcint-door002Domain:AAYl267UoWA] Unable to resolve IP address 10.42.38.112 to a canonical name
Enabling the pools external and internal IPV4/V6 counterparts in the /etc/hosts file allows root and roots transfers to work.
gfal-copy -f /etc/services roots://dcint-door002.sdcc.bnl.gov:1096/pnfs/usatlas.bnl.gov/cgamboa/test.root.write.cern.secure.1
Copying file:///etc/services [DONE] after 3s
For DAVs this is not necessary. Transfers worked without issue.
The text was updated successfully, but these errors were encountered:
I looked at your issue a little more carefully and realized this is not what I initially thought. In our case on dual stack pool nodes that resolve both IPv4 and IPv6 we had to add explicit name resolution to /etc/hosts on these pools but not on the door.
In your case you are adding pool name resolution to the door /etc/host. That's not right.
The resolution via DNS is enabled on doors and pools resolve to internal (IPV4/IPV6) components. Currently the production system will prefer using the internal LAN to communicate to all elements on the system.
This is why the internal components are resolve via DNS
Dear all,
Currently we are testing dCache doors with standard pool redirection for transfers.
I noticed that unlike with DAVs I must make sure the pool’s IPs must be on /etc/hosts at the door to be able to make the transfer work.
Why I do not have to do this for DAVs to work?
See below:
For xrootd doors tests external and internal transfers need the door to be able to resolve to external IP address pair of the pool or the internal depending on the clients network locality the:
Example external client access via xroot
Door logs
At the door:
The pool's DNS resolution is at the door is:
The xroot door needed to be reloaded
At the door,
Enabling the pools external and internal IPV4/V6 counterparts in the /etc/hosts file allows root and roots transfers to work.
For DAVs this is not necessary. Transfers worked without issue.
The text was updated successfully, but these errors were encountered: