jlo@elan.UUCP (Jeff Lo) (11/04/88)
Lately we have been getting a number of "out of gnodes" or "gnode: table is full" messages. The machine is a MicroVAX II running Ultrix 2.3. The machine is both an NFS server and client with two other machines (a Sun and an HP) on the network. I am assuming from looking at include files, etc., that is is a file system thing. It hasn't caused any problems yet that we have seen, but I would like to know what is going on. Thanks! -- Jeff Lo ..!{ames,hplabs,uunet}!elan!jlo Elan Computer Group, Inc. (415) 322-2450
mark@drd.UUCP (Mark Lawrence) (11/05/88)
jlo@elan.UUCP (Jeff Lo) wrote: %% Lately we have been getting a number of "out of gnodes" or "gnode: table %% is full" messages. ... hmmm -- I wasn't aware that DEC Ultrix was utilising FSF's implementation of the popular virtual file system. I wonder if RMS is aware? :-) Mark
srinivas@speedy.cs.wisc.edu (V. Srinivasan) (11/05/88)
In article <374@elan.UUCP|, jlo@elan.UUCP (Jeff Lo) writes: | Lately we have been getting a number of "out of gnodes" or "gnode: table | is full" messages. The machine is a MicroVAX II running Ultrix 2.3. The A gnode is a generic i-node like structure created in memory for accessing local and remote files in a uniform way. Every open file in the kernel has one gnode structure associated with it. The gnode is the Ultrix counterpart of the old in-core i-node. Gnodes are allocated from table with a static size fixed at kernel compile time. In your case the problem seems to be that too many files are being opened at the same time and the system runs out of gnodes. | Jeff Lo | ..!{ames,hplabs,uunet}!elan!jlo | Elan Computer Group, Inc. | (415) 322-2450 --Srini (srini@cs.wisc.edu)
tp@granite.dec.com (t patterson) (11/06/88)
In article <374@elan.UUCP> jlo@elan.UUCP (Jeff Lo) writes: >Lately we have been getting a number of "out of gnodes" or "gnode: table >is full" messages. The machine is a MicroVAX II running Ultrix 2.3. Somebody else has already replied & said what gnodes are, so I'll just babble about what to do when you run out of them... (I don't believe this is documented in the Ultrix "Guide to System Configuration File Maintenance") first, do "/etc/pstat -T" -- this ought to give you an idea of the size of some of your static tables and how close you are to overflowing them. (of course, you only see the current state of the tables.) anyway, until somebody releases Ultrix with dynamically expanding system tables, you'll have to make yourself a new kernel if you wish to avoid these messages. do "cd /usr/sys/conf" if you poke around in param.c, you should see: #define NGNODE ((NPROC + 16 + MAXUSERS) + 32 + NTEXT) you want NGNODE to be increased. you might want to get serious and try and determine which of these tables your system is pounding on so that you can increase just the ones that need it... I am usually really lazy -- in the kernel config file, I'll just put in something like "maxusers 32" on a uVAX and if you read the #define's in param.c, you'll see that this pretty much increases the size of everything... once you'd decided how you want to twiddle NGNODE and the other values, make and install a new kernel. a lot of work just to increase a couple little tables... -- t patterson internet: tp@decwrl.dec.com uucp: decwrl!tp disclaimer: this message represents no official positions of DEC