I have installed the 1.7 Hosted PageTest and I have successfully pulled the latest tip from SVN. My configuration is such that I placed the “site” on physical hardware, but I am running the URLblast.exe on virtual machine (VMware Fusion in the event it is relevant). The site runs on port 8000.
I have observed the following… URLblast starts clean and reports no errors (even with debug=1) and is able to collect work and attempts to run the job, but it would appear the browser window(s) is opened for only fractions of second before it gone again. I briefly see the title bar reflect the test URL before the window is gone (and IE cannot display… I cannot see it is too fast). I am NOT running on PORT 80 and instead port 8000. The server never receives the results and nothing else appears in the logs other than the fetch attempt for additional work. My log data follows below. Please let me know if you have any additional suggestions. Is it possible that the VM clock tic is creating havoc? I have tried increasing the timeout but it does not seem to help at all. I appreciate any advice.
2010/03/02 12:04:08 -1 2 1 0 0 URLBlast Started
2010/03/02 12:04:08 -1 2 1 4 0 CPU String Intel(R) Core(TM)2 Duo CPU T9300 @ 2.50GHz
2010/03/02 12:04:08 -1 2 1 5 2493 CPU MHz
2010/03/02 12:04:08 -1 2 1 6 511 Total RAM (MB)
2010/03/02 12:04:08 -1 2 1 7 30710 Disk Size (MB)
2010/03/02 12:04:08 -1 2 1 8 24935 Disk Free (MB)
2010/03/02 12:04:08 -1 2 1 9 5 OS Version Windows XP Service Pack 3 (5.1.2600 SP 3.0 Suite 0x00000100 Product 1)
2010/03/02 12:04:08 -1 2 1 10 0 Computer Name XXXXXXXXX
2010/03/02 12:04:08 -1 2 1 11 8 IE Version 8.0.6001.18702
2010/03/02 12:04:08 -1 2 1 12 217 URLBlast Version 2.0.0.217
2010/03/02 12:04:09 -1 2 1 3 0 Browser Launched Cleared Cache-Run_1^http://www.google.com
2010/03/02 12:04:22 -1 2 1 3 0 Browser Launched Cached-Run_1^http://www.google.com
Urlblast has problems if the server is running on anything other than 80. There were 2 bugs, one is fixed in the tip of the code and the second one has yet to be fixed. Until it is fixed there is no way to get it to work correctly on a different port. If it is critical too have it working on a different port I can see about fixing it in the next couple of days.
Thank you… No it is not critical to correct the bug immediately. However, I seem to still have a problem which I observed, but could not confirm until I had the port issue solved (see below).
One additional question:
Is the rapid window closure normal for win32 binary? Where is the data stored prior to being transmitted to the server or is it simply held in memory and dumped?
Creating the iptables redirect solved this issue (there is no rush to deliver a fix) but there is definitely something wrong with timing methods on a virtual machine (XP) because I can assure you (I am watching the running VM on my Mac live) that the execution time was <2-3 seconds to open and then close the window. IE works fine when run manually and there are no browser errors. However, the test server is reporting that it timed out after a negative time duration. Suggestions? How are the current publicly available pagetest test nodes provisioned? As VMs or dedicated hardware? The screenshot shows Internet Explorer cannot display the web page… The results are below:
Error loading page: -2146697213 (0x800C0003)
Page load time: 523.956 seconds
Time to first byte: 0.000 seconds
Time to Start Render: 525.548 seconds
Time to Document Complete: 523.956 seconds
Time to Fully Loaded: 523.956 seconds
Bytes sent out: 0.000 KB
Bytes received: 0.000 KB
DNS Lookups: 0
Connections: 0
Requests: 0
OK Requests: 0
Redirects: 0
Not Modified: 0
Not Found: 0
Other: 0[/code]
All of the public instances (and everything we run at AOL) is on physical hardware because the VM’s introduce timing variance. That said, it should still WORK, just with less accurate timings so something else isn’t working. I do most of my dev testing in a VM environment.
Do you have a suggestion? My debug=1 on the XP instance is as follows below. The user running the on the VM is NOT the same user as the ‘tester’ user… Should the tester user be an admin as well? The running user is an administrator on the box. There is corporate antivirus on the XP instance. I have included the OS and IE info as well.
[code]
2010/03/02 16:43:14 -1 2 1 9 5 OS Version Windows XP Service Pack 3 (5.1.2600 SP 3.0 Suite 0x00000100 Product 1)
2010/03/02 16:43:14 -1 2 1 11 8 IE Version 8.0.6001.18702
2010/03/02 16:43:14 -1 2 1 12 217 URLBlast Version 2.0.0.217
2010/03/02 16:44:00 -1 2 1 20 0 Debug Failed to retrieve an url for testing.
2010/03/02 16:44:00 -1 2 1 20 0 Debug Requesting work from example.com/work/getwork.php?location=Test_Corporate_SF&key=TestKey123
2010/03/02 16:44:00 -1 2 1 20 0 Debug Url to be teested: http://www.google.com
2010/03/02 16:44:02 -1 2 1 20 0 Debug Launching… user=‘user1’, path=‘C:\Program Files\Internet Explorer\iexplore.exe’, command line=‘“C:\Program Files\Internet Explorer\iexplore.exe” about:blank’
Sorry, missed the additional questions. The binaries can be build with Visual Studio 2008. You should be able to just grab the tip, open the project and do a build. That said, it may be easier if I just cut a new build (unless you already have a windows dev environment)
Well you right about something else being wrong. I assumed and perhaps incorrectly that PageTest supported IE8, but this may not be true. When I try to run the PageTest plugin manually on the XP instance, I get the same and identical errors. If IE8 is supported and (and I am running the latest PageTest dll) is there any way to enable additional debugging of a manual instance of PageTest plugin?
What version of the plugin are you running (help-about)? There were a few releases after the one bundled in the 1.7 zip file so odds are you’re not running the latest.
I appear to have solved the issue… Do you have an unpublished dependency on a particular version of .NET? I do not see anything listed in the pagetest wiki suggesting that there is a dependency there (I did not search all that hard), but I performed a windows update and installed the latest patches as well as the latest .NET 3.0 SP and suddenly pagetest as a plugin is working when run manually or via urlblast.
Version is 2.0.0.217 which was bundled in 1.7 which appears to be the same version as available from sourceforge. I copied the dll from the msi (after installing it) to the C:\Web Site directory.
i’m getting the exact same error message/log output on my instance (Windows XP/ IE8). I think this is a physical box, not a VM, i’m on it using remote desktop.
When disabling the CIEhook i can at least use the browser, with this addon active nothing works - it seems.
Do you have any antivirus installed? If so, which one? I have seen conflicts with the “web shield” part of some antivirus solutions.
If you are having issues with getting IE working with the browser plugin you can uncomment the browsrExe=pagetest.exe line in loactions.ini on the server and it will switche to use a dedicated browser (with the IE engine). The main difference is that it doesn’t load any other plugins that you have installed so it avoids any compatibility issues.
That did it! Now i can at see that its working and i get summarized results. The waterfall is empty though. There is no content breakdown, but the screenshots are there. I’m using php as a module on windows apache, maybe there is a problem with the gd module…
Are the screen shot thumbnails there? If so then it’s probably not a gd problem (gd is used to generate those as well). Can you check the error.log on the apache server to see if there is anything useful in there? Maybe a memory limit being hit or something.
