hdtodd@eagle.wesleyan.edu (04/03/91)
I am running two diskless 2500's from a disked 2500 over Ethernet and via a bridge. The bridge is PCBRIDGE, an AT&T 6300 with two WD8003 boards running Vance Morrison's public-domain PCBRIDGE code. When the 2500's were all side-by-side, performance of the diskless machines was fairly respectable ... booting was reasonably fast, etc. Now that they have the bridge between them (for security reasons), they require 20 minutes to boot and frequently hang in the process. It *SEEMS* clear that the problem is in the bridge. We suspect that the 8K buffer limit in the WD boards is a major problem, one that Morrison noted in conjunction with NFS support. One solution would seem to be to replace the bridge: not easy to do because of financial constraints. Another solution might be to tell netman to use small packets. I cannot find man pages or manual entries for netman parameters: are there any options? It appears that the diskless systems actually generate the requests and netman merely responds: if that's the case, is there any way to limit the packet sizes on diskless nodes as they boot? Thanks for any help you can offer. David Todd, Wesleyan University
mort@apollo.HP.COM (Stephen Moriarty) (04/09/91)
In article <1991Apr2.211328.41202@eagle.wesleyan.edu> hdtodd@eagle.wesleyan.edu writes: > Another solution might be to tell netman to use small packets. I >cannot find man pages or manual entries for netman parameters: are there any >options? It appears that the diskless systems actually generate the requests >and netman merely responds: if that's the case, is there any way to limit the >packet sizes on diskless nodes as they boot? The only option to netman is -db. When netman is run in a window, the -db option tells it to display more verbose information regarding the replies netman makes. As for regulation of packet size, netman gives replies of a size requested by its client. Therefore, as you suspect, it is netboot that must ask for smaller replies. This is not configurable. Note also, that it is the number of pages requested, not the packet size, that netboot specifies. Finally, as you are no doubt well aware, placing a bridge between a diskless node and its paging partner is not a supported configuration. ARPA: mort@apollo.hp.com UUCP: ...{decvax, umix, mit-eddie}!apollo!mort Apollo, a subsidiary of Hewlett Packard, 300 Apollo Drive, Chelmsford, MA. 01824 Argue for your limitations, and they are yours.