Huge sessionid.tdb file with dead records.

[Hemanth Thummala]

Hi All,

We are using samba 3.6.12+ stack. And recently facing the issue of
sessionid.tdb file growing very huge( > 1.6GB ). Here are two instances
where we have seen this problem.

First case we have seen tdb file grown to 1.6G

tdb> info
Size of file/data: 1618509824/159090
Number of records: 100
Smallest/average/largest keys: 14/14/15
Smallest/average/largest data: 1576/1576/1576
Smallest/average/largest padding: 401/401/402
Number of dead records: 1 ===>
Smallest/average/largest dead records: 1618305000/1618305000/1618305000 ==>
Number of free records: 2
Smallest/average/largest free records: 464/1228/1992
Number of hash chains: 131
Smallest/average/largest hash chains: 5/13/22
Number of uncoalesced records: 0
Smallest/average/largest uncoalesced runs: 0/0/0
Percentage keys/data/padding/free/dead/rechdrs&tailers/hashes:
0/0/0/0/100/0/0

Integrity check for this specific file is failed. But samba sessions were
running fine without any issues. But only concern was all smbd processes
mapping to this tdb virtual address space were showing 1.6G.
tdb> check
Dead space at 204800-1618509824 (of 1618509824)
Hashes do not match records
Integrity check for the opened database failed.

We had to restart samba to recover from this scenario. Actually I came
across a command "net servid wipedbs" which is supposed to cleanup all dead
records in sessionid.tdb. But this one also doesn't seems to be working in
this case.

Recently we have seen one more case for same sessionid.tdb file. But this
time it is 1.7G with very huge dead record.

tdb> info
Size of file/data: 1702981632/176576
Number of records: 111
Smallest/average/largest keys: 13/14/15
Smallest/average/largest data: 1576/1576/1576
Smallest/average/largest padding: 401/401/403
Number of dead records: 1 ===>
Smallest/average/largest dead records: 1702572008/1702572008/1702572008 ===>
Number of free records: 70
Smallest/average/largest free records: 464/2620/8040
Number of hash chains: 131
Smallest/average/largest hash chains: 0/2/6
Number of uncoalesced records: 22
Smallest/average/largest uncoalesced runs: 1/1/2
Percentage keys/data/padding/free/dead/rechdrs&tailers/hashes:
0/0/0/0/100/0/0
tdb>

I would like to check if anyone else has faced the same issue in past. I
have seen people complained about messages.tdb and locking.tdb.
Any thoughts why the dead records growing that huge? Would like to
understand before I go through code in detail.

Thanks,
Hemanth.