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

Re: [rhel6-branch 2/2] memory: increase the RAM limits, check for URL installs (#549653).



-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 05/05/2010 04:59 AM, Ales Kozumplik wrote:
> Due to higher memory consumption of LVM, kernel, anaconda and other
> running processes in general our current memory limits are insufficient
> and the increase is neccessary. I am also introducing additional check for
> URL installs which require another 108 MB from the ramdisk, no need to
> penalize other install methods.
> 
> This is the current memory budget:
> - unpacked initrd takes 63 MB
> - running processes (anaconda, loader, udev, nm): 67 MB (estimate)
> - peak LVM memory consumption: up to additional 56 MB
> - with URL install install.img takes additional: 108 MB
> - memory requriements of X additional: 256 MB (estimate)
> 
> Currently we always expect there will be calls to lvm (though we could do
> another memory check just at the moment where the necessity arises). All
> of the estimates are on the upper bound, still I am running into OOM
> easily with text install+http stage2+lvm detection with the old limits.
> ---
>  anaconda     |   37 +++++++++++++++++++++++++++++--------
>  isys/isys.c  |    1 +
>  isys/isys.h  |    5 +++--
>  isys/isys.py |    1 +
>  4 files changed, 34 insertions(+), 10 deletions(-)
> 
> diff --git a/anaconda b/anaconda
> index e85a70a..a5c586f 100755
> --- a/anaconda
> +++ b/anaconda
> @@ -358,24 +358,41 @@ def runVNC():
>                      sys.stdin.readline()
>                      iutil.execConsole()
>  
> -def checkMemory(opts):
> -    if iutil.memInstalled() < isys.MIN_RAM:
> +def check_memory(opts, display_mode=None):
> +    def within_available_memory(needed_ram):
> +        # kernel binary code estimate that is
> +        # not reported in MemTotal by /proc/meminfo:
> +        epsilon = 15360 # 15 MB
> +        return needed_ram < (iutil.memInstalled() + epsilon)

Putting methods inside methods really bugs me. I don't see any reason
for this.

Ack the rest. I think being more flexible/realistic on memory
requirements is a good thing.

- -- 
Brian C. Lane <bcl redhat com>
Red Hat / Port Orchard, WA
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)

iQEVAwUBS+Gx1BF+jBaO/jp/AQI4+Qf/ZraOz3k5ymQpD4bzYj3m56CnZt/RE0SE
J2oUOdoRHAJb/eeJDNc2tpgrYkN1JWbRdEJhVsyFBEsnf+YCOS0MUeWZgMd9L08o
+OzznHXWevgZ2LE1gxIpur/ofCjT1h0MrauNkL9LjjJfvwSXOK3RiT31R3/qQndn
a3Jua6TAOojKrroHFUl4q1ShUAxkx8Ytkz90Nng6AG8AlGh2AI/f7lETvxCQgmcS
CVvplqnjmwkDsmvRwejPkqbk6wimxn4YKxdGEmCZO5l+FF1+rhkKXuRjqJgMJW3F
Ius1td818xQjhZQpRyy6h228i083Rzelc7CtDSpglj8v35kav2hZBg==
=cnyD
-----END PGP SIGNATURE-----


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