-
Story
-
Resolution: Done
-
Minor
-
rhel-7.3, rhel-9.5.z
-
Low
-
rhel-fs
-
ssg_filesystems_storage_and_HA
-
1
-
False
-
False
-
-
None
-
None
-
None
-
None
-
If docs needed, set a value
-
-
All
-
None
-
57,005
Description of problem:
A DFS share can host the same data on multiple machines across a whole AD forest. When such a DFS share is accessed by mount.cifs, it seems that AD sites are not taken into account and the share is provided by an AD server from a random site.
For instance:
\\foo.com\share -> \\site1.foo.com\share
-> \\site2.foo.com\share
Reading through [1] the following should happen (at least this is what MS clients are supposed to do):
When a UNC is accessed, a list of servers hosting this share is returned. The calling client should then resolve those server names and send CLDAP pings to all those servers in weighted random order. The client will then send the mount request to the first server that responds to the ping request.
[1] https://msdn.microsoft.com/en-us/library/cc717360.aspx
It seems that cifs.upcall (or some other component - not sure) is not sending those CLDAP pings to identity the server the client should talk to.
Version-Release number of selected component (if applicable):
cifs-utils-6.2-9.el7.x86_64.rpm
How reproducible:
Steps to Reproduce:
1.
2.
3.
Actual results:
Expected results:
Additional info:
- external trackers