Am I right in thinking that BOINC v6.2.18 still has the bug whereby once a file transfer (download in this case) has been attempted and failed, subsequent retries use the same cached IP address, instead of a fresh DNS lookup?

SETI has one download server address (boinc2.ssl.berkeley.edu), but two servers (208.68.240.13 and 208.68.240.18). Round-robin DNS should give you a random but equally-probable chance of connecting to either server.

Today, 208.68.240.13 has an HTTP service problem, so download requests have a 50% chance of hitting the bad server. But once the IP is cached, every retry stays bad: only restarting the BOINC service seems to get things started again.

And on one of my retries, I got

09/10/2008 17:10:47|SETI@home|[file_xfer] Started download of file 18au08af.30717.24612.15.8.186
09/10/2008 17:10:48||[http_debug] [ID#4] info: About to connect() to boinc2.ssl.berkeley.edu port 80 (#0)
09/10/2008 17:10:48||[http_debug] [ID#4] info:   Trying [color=red]13.240.68.208[/color]... 
09/10/2008 17:11:09||[http_debug] [ID#4] info: Timed out
09/10/2008 17:11:09||[http_debug] [ID#4] info: Failed connect to boinc2.ssl.berkeley.edu:80; No error

(note the reverse-order IP address) - but that was on BOINC v5.10.13, and I know you don't like to talk about old versions here.....

Found a second machine trying to contact the reverse IP address:

09/10/2008 23:59:41|SETI@home|[file_xfer] Started download of file 18au08ag.30754.18068.12.8.90
09/10/2008 23:59:41||[file_xfer_debug] PERS_FILE_XFER::start_xfer(): URL: http://boinc2.ssl.berkeley.edu/sah/download_fanout/2d1/18au08ag.30754.18068.12.8.90
09/10/2008 23:59:42||[http_debug] [ID#0] info: About to connect() to boinc2.ssl.berkeley.edu port 80 (#0)
09/10/2008 23:59:42||[http_debug] [ID#0] info:   Trying [color=red]13.240.68.208[/color]... 
10/10/2008 00:00:03||[http_debug] [ID#0] info: Timed out
10/10/2008 00:00:03||[http_debug] [ID#0] info: Failed connect to boinc2.ssl.berkeley.edu:80; No error
10/10/2008 00:00:03||[http_debug] [ID#0] info: Expire cleared
10/10/2008 00:00:03||[http_debug] [ID#0] info: Closing connection #0
10/10/2008 00:00:03||[http_debug] HTTP error: couldn't connect to server

This one was Windows XP, previous one Windows Vista - but still BOINC v5.10.13

Could someone check this out with a v6.2.xx build, please, next time there's an extended download server outage? (I can't test against 208.68.240.13 again, because Matt has kicked it, and it's working now).

Edit: And a Windows 2000 machine:

2008-10-10 00:18:45 [SETI@home] [file_xfer] Started download of file 19au08ac.10227.8252.6.8.64
2008-10-10 00:18:45 [---] [http_debug] [ID#0] info: About to connect() to boinc2.ssl.berkeley.edu port 80 (#0)
2008-10-10 00:18:45 [---] [http_debug] [ID#0] info:   Trying [color=red]18.240.68.208[/color]... 
2008-10-10 00:19:08 [---] [http_debug] [ID#0] info: Timed out
2008-10-10 00:19:08 [---] [http_debug] [ID#0] info: Failed connect to boinc2.ssl.berkeley.edu:80; No error
2008-10-10 00:19:08 [---] [http_debug] [ID#0] info: Expire cleared
2008-10-10 00:19:08 [---] [http_debug] [ID#0] info: Closing connection #0
2008-10-10 00:19:08 [---] [http_debug] HTTP error: couldn't connect to server

Note that this was trying to contact the (reverse of the) server that didn't have problems. Maybe we don't need an outage to test, just some congestion? That should be easy enough to organise at SETI....