Coda File System

Re: ERROR: rds_zap_heap RVM_EINTERNAL message.

From: Jan Harkes <>
Date: Mon, 24 Feb 2003 03:13:43 -0500
On Sun, Feb 23, 2003 at 11:14:16PM -0700, Carl Baldwin wrote:
> //  I ran vice-setup and everything
> seemed to go well until right at the end when it gave this error:

Hmm, interesting. I get the same error with a 200M data file. The 130M
one works. One of those magic numbers is probably wrong.

... strace'd it ...

old_mmap(0x50000000, 184549376, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = -1 ENOMEM (Cannot allocate memory)

It looks like we can't mmap the data file at the address we want it
even. But we did set ulimit -d unlimited. Ok found the problem, for some
reason the 2.4 linux kernels are disabling overcommit, and we're
probably trying to map more than the system has available. Enabling
overcommit (echo 1 > /proc/sys/vm/overcommit) and the OOM killer kicks
in because I really don't have enough memory + swap available.

Solution, add more swap or use a smaller data segment.

> What is your log partition? /var/log/codalog

btw. this is not a 'logfile' with human readable stuff. It is a binary
log of transactions that need to be applied to the data file. Similar to
a (hidden) ext3 journal file. Not that the location matters much, but
you might mess up your terminal if you are trying to read it.

Received on 2003-02-24 03:16:10