yum with recent soas builds on XO-1

Mikus Grinbergs mikus at bga.com
Fri Aug 14 07:55:39 EDT 2009


Martin Dengler wrote:
>> Not sure about *that* mountpoint in particular, but a forest of
>> "invisible" tmpfs mounts is present in the XO image because we use
>> "stateless" Fedora.
>>
>> This allows olpc-update to work its magic.
> 
> This thread segment is about SoaS now (a bit OT for devel@, I agree),
> which is not doing anything like the initrd /versions hardlinking.

I do not understand the intent of the sentence immediately above.

Recently on F11-on-XO1, *both* when I am running a soas-stream 
build, and when I am running a fedoraproject-stream build -- 
/etc/cache/yum ends up being mounted as a tmpfs.  So whether the 
mount is done "invisibly" (soas livesys) or "visibly" (fedora 
initrd) ought not affect the discussion of whether putting 
/etc/cache/yum onto a tmpfs is desirable for a memory-limited 
platform such as the XO-1.

The comment in livesys says this is being done to save writing to 
the overlay for caches.  [Yet another situation where "mother knows 
best" (i.e., configuration decisions are being made by the developer 
for the end user).]


To me the interesting thing is that on fedoraproject-builds yum 
(with its cache on tmpfs) works for me so far, while on SoaS3-builds 
yum (with its cache on tmpfs) often fails.  Yes, that may be about 
SoaS now -- but ought to serve as a caution to all F11-on-XO1 
developers that "independent" functions (such as 'yum') can end up 
seriously affected.


mikus




More information about the Devel mailing list