(no subject)

Jason Coene jason at dhnw.net
Tue May 1 18:07:10 GMT 2001


I have been tracking this bug as much as possible, and it appears there is a
severe problem in the code that deals with .tdb files. Lately our samba 2.2
testbed has been having painfully slow network responsiveness. It would go
30-60 seconds without a response to a client on many occasions (both
browsing the network, and opening files/refreshing directories). Clients
include Windows 98se, nt, and 2000. When it would "hang" the client
machine's worker thread due to not delivering a timely response, the samba
server would eat 80% cpu time per "working yet buggy" thread. This is on a
p3-600, 256mb ram, raid1 30gb running OpenBSD Kernel 2.8. Several times it
got so severe the client's explorer would die, and the client pc would have
to be restarted.

Today i found a 2.1 GB brlock.tdb file, a 1.3 GB connections.tdb, and a 1.6
GB locking.tdb file in my lock directory. Samba team, are you aware of this
yet? Searching the mailing lists quickly, it seems one other person has
reported it.


Jason Coene

More information about the samba-technical mailing list