michaelw@syacus.acus.oz (Michael Wang) (08/09/90)
Dear all, I am trying to install ISODE-6.0 on a Sun using an NFS mounted disk. The 'make' is taking exceedingly long time to complete (it has been 3 days now and is still crawling along). I suspect this has got something to do with NFS. I need some suggestions! Our Configuration: a) The Server: a Unisys 6050 (Convergent Technology CTIX) AT&T V.3 4 nfsd and 4 biod b) The Client: a Sun 3/60 SunOS 3.5 4 nfsd and 4 biod Server's file system is mounted with timeo = 30; retrans = 6 (we started to use the default timeo=7 and retrans=3 and that was BAD!) Our Situation: I am installing ISODE-6.0 on the client and due to available disk space I have to install ISODE-6.0 on the server's disk. I have observed the following: 1) There is a massive number of retrans/timeouts on the client's side (approx. 10%). 2) From the client, if I do a 'ls' of the directory that the 'make' is currently doing work in then I won't get any response for 5-10 minutes. In fact, I cannot even kill it from another terminal. According to 'ps' the job is in a Disk wait with PRIority -5 which means it is in a non-interruptible wait. (This happens a lot!! Extremely annoying!) 3) From the client, a long wait will occur for the following type of jobs. a) 'ar' of LARGE number of object files resident on the server's disk and target the resulting library to the server's disk; b) 'ranlib' on those libraries; c) 'cp' of any file whose size exceeding 0.3 Mb (this is a rough guess) from the server's disk to a different directory on the server's disk. The 'cp' executed on the server only takes about 13 seconds for a 1M file whereas it could take up to 2 hours (!!!) to 'cp' a file from the client. Again, these jobs will appear to 'ps' to be in a Disk wait with PRIority -5. 4) From the client, "NFS not responding still trying" message would come up but I can still 'ls' directories on the server or even edit files resident on the server. (Sometimes, that message does mean what it means though.) Our questions: 1) Is there anyone with a similar problem? 2) Would increasing the number of daemons help? (Mind you no one else is using NFS apart from my 'make'.) 3) Can this problem be fixed or do I have to sit here and suffer ?! Thank you all.