Phantom GFE's on NetWare 5 with NLM 5 (Networking Products)
At 01 DEC 1999 08:44:27AM Dan Reese wrote:
We have one customer site that has had two phantom GFE's in the last two weeks using AREV 3.12, NetWare 5.0, and the Revelation NLM for NetWare 5. They are phantom GFE's in that they go-away if you down the server and bring it back up. The file that is having trouble is about 780MB. This behavior is similar to the behavior of NetWare 4.x without the turbodis patch and 3.x without fatfix.
The workstations in use are all Windows 95 with the Novell Client 3.1 (no service packs). We have done all of the standard setup things, like setting up REVPARAM files, verifying that file handles contain the F's, disabling all workstation level caching including Novell Client caching, we have forced frame types to 802.2 and set IPX as the first protocol, we don't use MAP ROOT drives, etc.
Has anyone else seen phantom GFE's on NetWare 5?
Does anyone have any ideas about what might be going on?
At 02 DEC 1999 05:54AM Steve Smith wrote:
Dan,
You've got me guessing. Check the server for IPX retries or checksum errors (I'm thinking of reflections of signal along the network segment). If it looks clean then you might try adding 802.3 as a protocol, and see if this helps. Also, restrict the account to 1 gB free space on the AREV volume, regardless of what is physically present. Turn all transaction tracking off for the AREV volume. Check that there is no virus scanning software present,or disable it for *.OV file suffixes. Increase the parameters for network buffers and file handles, if possible. Ensure that these is sufficient disk space for temporary files, wherever these are created.
Steve
At 02 DEC 1999 06:30PM Dan Reese wrote:
Why the restriction to 1GB?
Is this 1GB in addition to the data that is present?
Dan
At 06 DEC 1999 02:53AM Steve Smith wrote:
Dan,
Free space. There's 2.1gB limit (2^31 bytes) to what DOS can report back as free space. More than this, and the value goes to a signed double integer (bit 32 set), and becomes negative (which gets interpreted as out of space). I've seen quirky AREV behaviour without this limit set. 1 gB was being conservative.
Steve