[PATCH][XM-TEST] Test for bug #533

[PATCH][XM-TEST] Test for bug #533

[Dan Smith]

[-- Attachment #1: Type: text/plain, Size: 1192 bytes --]

This patch adds a test that attempts to create and destroy a domain
1025 times.  Currently, this causes xm to crash after the 1014th cycle
with the following stack trace:

> Traceback (most recent call last):
>   File "/usr/sbin/xm", line 8, in ?
>     from xen.xm import main
>   File "/usr/lib/python/xen/xm/main.py", line 34, in ?
>     import xen.xend.XendProtocol
>   File "/usr/lib/python/xen/xend/XendProtocol.py", line 20, in ?
>     import httplib
>   File "/usr/lib/python2.4/httplib.py", line 70, in ?
>     import mimetools
>   File "/usr/lib/python2.4/mimetools.py", line 6, in ?
>     import tempfile
>   File "/usr/lib/python2.4/tempfile.py", line 33, in ?
>     from random import Random as _Random
>   File "/usr/lib/python2.4/random.py", line 44, in ?
>     from math import log as _log, exp as _exp, pi as _pi, e as _e
> ImportError: /usr/lib/python2.4/lib-dynload/mathmodule.so: cannot open
> shared object file: Too many open files

Since this test takes a long time to run, I have not enabled it by
default, but it should be included in the tree to make bug #533 easily
reproducible.

Perhaps xm-test should have a configure option that enables long,
exhaustive tests?


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: massive_create.patch --]
[-- Type: text/x-patch, Size: 1021 bytes --]

# HG changeset patch
# User dan@localhost.localdomain
# Node ID 3e84a0ac070b4c85c81d593b70ee9a23e0c9b2cb
# Parent  fcc833cbaf827327ad9e9348c93aaf7f99f0253f
Add a test for ~1000 create/destroy cycles.  Currently, this causes
xm to crash.

Signed-off-by: Dan Smith <danms@us.ibm.com>

diff -r fcc833cbaf82 -r 3e84a0ac070b tools/xm-test/tests/create/16_create_massive.py
--- /dev/null	Thu Jan  1 00:00:00 1970 +0000
+++ b/tools/xm-test/tests/create/16_create_massive.py	Tue Feb 14 07:31:03 2006 -0800
@@ -0,0 +1,21 @@
+#!/usr/bin/python
+
+# Copyright (C) International Business Machines Corp., 2005
+# Author: Dan Smith <danms@us.ibm.com>
+
+from XmTestLib import *
+
+COUNT=1025
+
+for i in range(0,COUNT):
+	print "Creating %i of %i..." % (i, COUNT)
+	try:
+		dom = XmTestDomain(name="massive-%i" % i)
+		dom.start()
+		console = XmConsole(dom.getName())
+	except DomainError, e1:
+		FAIL("Failed to start domain %i" % i)
+	except ConsoleError, e2:
+		FAIL("Failed to attach console to domain %i" % i)
+	
+	dom.destroy()

[-- Attachment #3: Type: text/plain, Size: 87 bytes --]

-- 
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms@us.ibm.com

[-- Attachment #4: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: [PATCH][XM-TEST] Test for bug #533

[Vincent Hanquez]

On Tue, Feb 14, 2006 at 07:34:24AM -0800, Dan Smith wrote:
> This patch adds a test that attempts to create and destroy a domain
> 1025 times.  Currently, this causes xm to crash after the 1014th cycle
> with the following stack trace:

The test is bogus and so if the bug.
seem you don't close the console fd by calling closeConsole at each loop.

> > shared object file: Too many open files

that it is basic ulimit behavior (usually 1024 is default configuration)

> +# Copyright (C) International Business Machines Corp., 2005
> +# Author: Dan Smith <danms@us.ibm.com>
> +
> +from XmTestLib import *
> +
> +COUNT=1025
> +
> +for i in range(0,COUNT):
> +	print "Creating %i of %i..." % (i, COUNT)
> +	try:
> +		dom = XmTestDomain(name="massive-%i" % i)
> +		dom.start()
> +		console = XmConsole(dom.getName())
> +	except DomainError, e1:
> +		FAIL("Failed to start domain %i" % i)
> +	except ConsoleError, e2:
> +		FAIL("Failed to attach console to domain %i" % i)
try to add here:
  +	console.closeConsole()
> +     dom.destroy()

and report back if it doesn't fix the problem.

[resent: forgot to CC xen-devel]
-- 
Vincent Hanquez

Re: [PATCH][XM-TEST] Test for bug #533

[Ewan Mellor]

On Tue, Feb 14, 2006 at 07:34:24AM -0800, Dan Smith wrote:

> This patch adds a test that attempts to create and destroy a domain
> 1025 times.  Currently, this causes xm to crash after the 1014th cycle
> with the following stack trace:
>
> > shared object file: Too many open files

Thanks Dan.  I'll apply your patch soon.  Meanwhile, could you run this
test and then send the output of lsof when you're a long way in (but
before it crashes).  It should be pretty easy to figure out where you're
leaking that file descriptor from that.

> Since this test takes a long time to run, I have not enabled it by
> default, but it should be included in the tree to make bug #533 easily
> reproducible.
> 
> Perhaps xm-test should have a configure option that enables long,
> exhaustive tests?

I thought we already had two speeds of xm-test.  Do you mean add a
third?  If so, that's fine, and a patch would be welcome!

Ewan.

Re: [PATCH][XM-TEST] Test for bug #533

[Dan Smith]

VH> The test is bogus and so if the bug.  seem you don't close the
VH> console fd by calling closeConsole at each loop.

You're right.  I think what is happening is that the test exhausts all
the fd's and then exec's xm with none free.

I was trying to replicate a situation reported on xen-users, so when I
hit this, I filed the bug and sent the test.  Turns out, I wasn't even
replicating the right situation anyway.

I'll write a test for the /actual/ reported issue and submit that if
the behavior is reproducible.

Just ignore me for now ... ;)

-- 
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms@us.ibm.com