Kernel param. for maximum size of single file in memory?

[Consul]

You know, I had the answer to this once upon a time, and now I can't find it anywhere.

We're running a MySQL cluster (version 5.0.18), and we're trying to convert some of our machines over to Gentoo from Slackware. At this point, we are trying to replace one of our cluster nodes with a Gentoo machine, and we're running into a configuration problem.

MySQL has to have the maximum file size ceiling (for a file in memory, not on the disk) set to a high enough number to hold all of the databases we store in it. On the three other machines, that's 768MB. Now, the MySQL config file itself (for NDB) has a setting for this, but it can't exceed what the kernel has set for it (ie, it's a soft-limit within the program itself).

I know there is a kernel parameter somewhere that adjusts this, and I even knew its name once, but I'll be darned if nearly two hours of Google searching on just about every permutation I can think of turned up nothing. (I did find something called SHMMAX, which is the max. size of a shared memory segment. I'm pretty sure that's not what I need, although we may need to increase that as well.) Any chance anyone here can steer me in the right direction?

I hope I described this problem clearly enough. Thank you all for your help.

[anonybosh]

Could it have something to do with the /etc/limits file (just a guess)?

Code: Select all

$ man limits

       The  limits  file  (/etc/limits  by default or LIMITS_FILE defined con- 
       fig.h) describes the resource limits you wish to impose. It  should  be 
       owned by root and readable by root account only. 

       By default no quotas are imposed on 'root'. In fact, there is no way to 
       impose limits via this procedure to root-equiv accounts (accounts  with 
       UID 0). 

       Each line describes a limit for a user in the form: 

       user LIMITS_STRING 

       The  LIMITS_STRING  is a string of a concatenated list of resource lim- 
       its.  Each limit consists of a letter identifier followed by a  numeri- 
       cal limit. 

       The valid identifiers are: 

       A: max address space (KB) 
       C: max core file size (KB) 
       D: max data size (KB) 
       F: maximum filesize (KB) 
   --->M: max locked-in-memory address space (KB) 
       N: max number of open files 
       R: max resident set size (KB) 
       S: max stack size (KB) 
       T: max CPU time (MIN) 
       U: max number of processes 
       K: file creation mask, set by umask(2). 
       L: max number of logins for this user 
       P: process priority, set by setpriority(2).

[ctford0]

i'm not sure what you mean... are you talking about how much memory a single process can use? i personally have used upwards of 3-4 gigs of memory (physical/swap combo) with no problems at all for one process. with no special settings

chris

[Consul]

Yeah, to tell you the truth, I'm not sure what I need myself anymore. I just remember reading once that there was a kernel setting that needed to be adjusted (it required a recompile of the kernel) that may solve this MySQL Cluster problem we're having. We actually looked at the limits file, and couldn't get anything to work.

Argh, it's just been so frustrating, you know?

I'm going to go ahead and take this subject to the appropriate for MySQL stuff, with a lot more detail. Thank you all for your help.