Webserver Downtime

From: Arglex126 May 2010 13:52
To: ALL1 of 83

Hey guys,

We are going to take the webserver down to move to the new gear on Thursday (tomorrow) starting at 12:00am (friday morning). It could be down for up to 2 hours.

From: _ (V1M)26 May 2010 14:05
To: Arglex1 2 of 83
I can get drunk if that helps.





Or even if it doesn't.
From: Arglex126 May 2010 14:06
To: _ (V1M) 3 of 83
Alcohol and servers don't mix well.
From: _ (V1M)26 May 2010 14:12
To: Arglex1 4 of 83
But alcohol and mixers serve well!
From: Feed the (ZUDZUG)26 May 2010 14:27
To: _ (V1M) 5 of 83
alcohol and serveuses too!
From: Gawd Almighty (WHITEKNAVE)27 May 2010 07:18
To: Feed the (ZUDZUG) 6 of 83
alcohol and cervices (cervixes).
From: Kune Dog (LIGTYMN)30 May 2010 18:05
To: Arglex1 7 of 83
The cut-off-image problem is happening worse than ever now.

http://www.pr0kforums.com/index.php?webtag=PR0K&msg=30114.11

Less than half of the images I posted today are loading fully.

http://www.pr0kforums.com/index.php?webtag=PR0K&msg=2252.3853
From: Jacfascist (JACFLASH)30 May 2010 19:20
To: Kune Dog (LIGTYMN) 8 of 83
Yeah, probably half of what you posted today didn't load for me - on the GFU's laptop, running IE8. Haven't checked at the house.
From: Kune Dog (LIGTYMN)30 May 2010 20:00
To: Jacfascist (JACFLASH) 9 of 83
I tried both Opera and FF (and FF in Linux).
From: Arglex130 May 2010 22:02
To: Kune Dog (LIGTYMN) 10 of 83
The server has not been upgraded yet. We had a refusal for downtime so we have to wait
From: lcek48 (LECK48)31 May 2010 03:12
To: Kune Dog (LIGTYMN) 11 of 83
Everything loaded just fine for me.
From: lcek48 (LECK48)31 May 2010 03:18
To: Kune Dog (LIGTYMN) 12 of 83
Ok they all loaded but for some reason when I load that thread I get logged out and have to relogin, weird.
From: Jacfascist (JACFLASH)31 May 2010 11:18
To: lcek48 (LECK48) 13 of 83
Did not load completely for me at home, WXP + Chrome. I'll try it in linux later.
From: Jacfascist (JACFLASH)31 May 2010 11:18
To: Kune Dog (LIGTYMN) 14 of 83
You still on Bellsouth/ATT?

Edit: nevermind, GFU is on Comcast, and they didn't load correctly at her place either.
From: Jacfascist (JACFLASH)31 May 2010 11:25
To: Kune Dog (LIGTYMN) 15 of 83
20 of them did not load here at work, and I was apparantly logged out while viewing them, in between this post and the prior.
From: Kune Dog (LIGTYMN)31 May 2010 12:58
To: Arglex1 16 of 83
The server is (sometimes) sending a premature "200 OK" HTTP response in the middle of transmitting an image. I say "sometimes" because an hour ago, everything was working perfectly and they all loaded fully, but ATM most of them are cut off again.

I ran Wireshark and captured an image which only partially loaded:


The 10.X address is my machine and 208.X is the server.

The client makes an HTTP GET request for the jpg file. Starting at frame 64, the server sends the binary file in chunks, which are ACKed by the client. Wireshark knows what's going on and calls these "TCP segments of a reassembled PDU," meaning that a higher-level protocol, http, knows about these pieces and will put them together.

That's all that happens until frame 150:


Now, seemingly out of nowhere, the server sends an HTTP "200 OK," indicating that the HTTP request completed sucessfully. At this point, the client's browser (Opera) displays the partially-assembled image and sends a reset packet (RST) to terminate the connection.

But then the server sends the rest of the image. Wireshark calls these "Continuation or non-HTTP traffic"; it doesn't know what they're for anymore, seeing as the server has already indicated the HTTP request is complete and instructed Opera to close the connection. Because the connection is closed, an RST is sent back every time the server sends more data.

I have reproduced this premature "200 OK" response several times with identical results. The only variation I've seen is when the client will send normal ACKs for the "Continuation" data after the "200 OK" response, instead of RST packets. After the server is finished, the client sends a single RST.

I also did a cap of a successful transfer. In this case, the server sends the "200 OK" when it should, when the entire image is transfered. The client sends a FIN response instead of an RST. This would indicate that the client does know something is wrong when it responds to an incomplete image with RST.
From: Kune Dog (LIGTYMN)31 May 2010 13:07
To: Kune Dog (LIGTYMN) 17 of 83
a couple notes:
  • Take care if you try to reproduce the http request above. It's one of the NSFW ones.
  • The server is cutting off (for me) the images in the above post.
From: Kune Dog (LIGTYMN)31 May 2010 14:59
To: ALL18 of 83
It's not anything specific to the Beehive page. I constructed two html files with all the pics (#2 is the NSFW ones) and they're cut off even when operned in isolation.
From: Jacfascist (JACFLASH)31 May 2010 15:12
To: Kune Dog (LIGTYMN) 19 of 83
49 incompletely loaded images in that first file for me. Mint linux, Chromium, though that info is probably extraneous now.
From: Jacfascist (JACFLASH)31 May 2010 15:18
To: Kune Dog (LIGTYMN) 20 of 83

41 in the second file, and I was logged out of the forum while viewing.

Though I have to say that the broken pictures didn't reduce my appreciation for the collection much :)