Bug 435271

Created attachment 296209 [details]
anaconda.log from the installed system

13:42:51 CRITICAL: Traceback (most recent call first):
  File "/usr/lib/anaconda/iutil.py", line 50, in execWithRedirect
    raise RuntimeError, "Error running " + command + ": " + msg
  File "/usr/lib/anaconda/lvm.py", line 84, in vgactivate
    stderr = output, searchPath = 1)
  File "/usr/lib/anaconda/partitions.py", line 1360, in doMetaDeletes
    lvm.vgactivate()
  File "/usr/lib/anaconda/packages.py", line 144, in turnOnFilesystems
    anaconda.id.partitions.doMetaDeletes(anaconda.id.diskset)
  File "/usr/lib/anaconda/dispatch.py", line 201, in moveStep
    rc = stepFunc(self.anaconda)
  File "/usr/lib/anaconda/dispatch.py", line 124, in gotoNext
    self.moveStep()
  File "/usr/lib/anaconda/gui.py", line 1018, in nextClicked
    self.anaconda.dispatch.gotoNext()
  File "/usr/lib/anaconda/iw/progress_gui.py", line 243, in renderCallback
    self.intf.icw.nextClicked()
  File "/usr/lib/anaconda/gui.py", line 1045, in handleRenderCallback
    self.currentWindow.renderCallback()
RuntimeError: Error running lvm: Cannot allocate memory

HVM guest is started with 512 MB of memory, newly created disk image.
The same configuration works when booting with boot.iso and http install method.

Additional info to match summary:
anaconda needs to show a nice dialog to the user informing them that available
memory is not enough but not produce a runtime traceback.
This needs to match system requirements for the related release and chosen
install methods.

How much memory does the machine have?  The minimums should be reasonable across
all platforms.  And usually, the bug is a red herring for something else (worse)
going wrong.

Jeremy,
the original bug was filed against a Xen HVM guest with 512 MB of memory (and no
higher limit IIRC).

512 megs should be enough -- if not, then that's a bug which should be fixed. 
Not papered over with a dialog.

Steps to reproduce:
1) boot from DVD (ia64) (xen hvm, 512 MB memory)
2) start VNC install
3) choose default install
4) select "Customize packages"
5) choose all package groups

When anaconda tries to format partitions the traceback happens.

Tested with RHEL 5.1 GA under the same memory limitations and steps to
reproduce. Can't see the traceback.

Also seeing this with 5.2 snap #4 steps to reproduce:
1) Create a new Xen HVM guest (use single file as storage - 15GB, 4GB memory)
2) Use kickstart that contains @everything
3) Choose: remove all partitions and create default layout

Attaching debug info shortly.

Created attachment 301820 [details]
anacdump.txt

Trying with same disk image (Xen) but without using lvm I had the same error.
Anaconda is running lvm to detect and delete previous partitions before creating
the new ones, right?

Trying with newly created disk image and using native partitions I see the same
error.

More info on this one:
It appeared that HVM guests can't bolloon memory as PV ones (I didn't know that).
So in all tests above the guest has been running with the start up memmory
allocation (512MB). Bumping that up to 2GB worked around the problem.

The faulty component to blame here is most probably yum. Python objects on 64bit
arches use 4 times more memory than on 32bit. Before the error happens yum is
resolving all deps.

After talking to ddumas today we've decided to release note this issue:

<text>
When installing all packages available for Red Hat Enterprise Linux 5.2 on ia64
the installation may fail on machines with 512 MB of memory. To work around this
perform a base or default install and then install all additional packages.
</text>

Brock, 
what flags we need to get this in the 5.2 release notes?

After running the tests on bare metal for ia64, it looks like 6 of the 10
installs using 512 Mb (via mem=512M cmdline parameter) fail, while 4 succeed.  

ddomingo is cc'd on this bug, and I'll ping him about getting this for 5.2 ...

thanks Alex, Brock; adding to RHEL5.2 release notes under "Installation-Related
Notes":

<quote>
(ia64) If your system has 512MB of RAM, attempting to install Red Hat Enterprise
Linux 5.2 may fail. To prevent this, perform a base installation first and
install all other packages after the installation finishes.
</quote>

please advise (before April 15) if any further revisions are required. also,
FYR, the updated RHEL5.2 release notes mockup can be viewed here:
http://intranet.corp.redhat.com/ic/intranet/RHEL5u2relnotesmockup.html

Thanks, Don!

verified entry at:

http://intranet.corp.redhat.com/ic/intranet/RHEL5u2relnotesmockup.html

will move to VERIFIED when we see this version in the compose tree ...

Is this only reproducable on 64-bit machines doing Everything installs?

Looks like it's ia64 only. I wasn't able to reproduce on other 64bit arches

This is only affecting one platform and I haven't seen any evidence that it
happens outside of Everything (or similarly gigantic) package sets.  Also it's
likely this is not an anaconda bug, but instead is a bug in yum memory usage or
Python's 64 bit object size.  I doubt there's really anything we can do in
anaconda.  NAKing for those reasons.

Development Management has reviewed and declined this request.  You may appeal
this decision by reopening this request. 

This bug has been added to the the RHEL 5.3 Release Notes (as quoted in the release note field of this bug).

If any revisions are required, please set the "requires_release_notes" flag to "?" and edit the Release Notes field accordingly.

Added New Release Notes Contents.