RE: Memory usage and /proc/meminfo question

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

> Hello Li,
> 
> Thanks for your response.
> 
> On 23 jun 2008, at 04:00, Li Xiaodong wrote:
> > According to the information you give, my suggestions are as follows:
> > 1. Check if there are daemon processes named pdflush running in your
> > system.
> 
> If have 2 pdflush daemons running.
Which means it is not because of IO that occupies too many caches.

> > 2. If NOT,
> > 	2.1 turn the value of /proc/sys/vm/dirty_backgound_ratio to 5 or
> > less. 	2.2 turn the value of /proc/sys/vm/dirty_ratio to 5 or less.
> > 	2.3 turn the value of /proc/sys/vm/dirty_expire_centisecs to 5 or
> > less.
> > 	2.4 wait
> > 3. If it still not works,
> > 	3.1 turn the value of /proc/sys/vm/vfs_cache_pressure to 1000.
> > 	3.2 run a much-memory-occupy program.
> > 	3.3 wait
> 
> Ok, I increased vfs_cache_pressure from 100 to 1000 and created a large
> process.
> Apparently it freed up some memory by shrinking the cache. So now my
> MemFree has increased and Buffers and Cached are much smaller. But it
> has not really solved the larger issue.
How many MBs does your large process occupy?
Make this process take 80%-90% of your memory.

> Adding MemFree + Buffers + Cached + AnonPages + Mapped + Slab still
> ends up at only about 348464, so I'm still wondering where the other
> 60% of my physical memory is hanging out.
> 
> Why did you ask me to increase vfs_cache_pressure. Do you think
> something is wrong with the dentry cache?
Maybe too many inodes and dentries live in your slab.
And without memory pressure, system may just leave them alone.
By increasing vfs_cache_pressure and running 'large process', system will
reclaim that part of memory back.


> I think dcache memory should show up under SReclaimable, is that right?
> 
> Do you agree that MemFree + Buffers + Cached + AnonPages + Mapped +
> Slab should add up to total physical memory or am I missing something?
> 
> > 4. If it still not works, reboot your computer...
> 
> Sure, but that would be admitting defeat.
> 
> Thanks, Joris.
> 




--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ


[Index of Archives]     [Newbies FAQ]     [Linux Kernel Mentors]     [Linux Kernel Development]     [IETF Annouce]     [Git]     [Networking]     [Security]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux RAID]     [Linux SCSI]     [Linux ACPI]
  Powered by Linux