Message boards :
Server and website :
Slow downloads and --pers_retry_delay_max 60
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 30 Apr 13 Posts: 109 Credit: 3,977,737,860 RAC: 575,436 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
I've added the startup parameter --pers_retry_delay_max 60 to boinc.exe, but I don't see anything in the log on startup that confirms the parameter was executed or even read. Should there be some indication in the log? When a file hangs during download there are two delays. First, about 5min10sec delay before anything happens, and when that completes the thing that finally happens is start of the retry delay of an additional varying number of minutes. I've got that second retry delay down to 15 seconds or less and I think that was accomplished by the cc_config line: <http_transfer_timeout>15</http_transfer_timeout>. Then what is the boinc startup parameter --pers_ supposed to do? I'm still getting the 5min10sec pause for every hang. Thanks. |
BeyondSend message Joined: 23 Nov 08 Posts: 1112 Credit: 6,162,416,256 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Here's one of my messages about this: https://www.gpugrid.net/forum_thread.php?id=4399&nowrap=true#44782 You have to start boinc.exe via command line, such as: "D:\Program Files\BOINC\boinc.exe" --pers_retry_delay_max 60 I make a shortcut for Windows and put it in my startup folder. It works on my 11 boxes. I wouldn't go lower than 60 seconds on this parameter: <http_transfer_timeout>60</http_transfer_timeout> |
|
Send message Joined: 30 Apr 13 Posts: 109 Credit: 3,977,737,860 RAC: 575,436 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Thanks. I am starting boinc.exe at the command line with the --pers... parameter. That's how I got the startup log I referred to. My question again: Should there be evidence in the startup log that confirms the parameter was taken? |
BeyondSend message Joined: 23 Nov 08 Posts: 1112 Credit: 6,162,416,256 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
When a file hangs during download there are two delays. First, about 5min10sec delay before anything happens, and when that completes the thing that finally happens is start of the retry delay of an additional varying number of minutes. I find no reference to the --pers parameter in the log file. However there should be no increasing delay on multiple retrys. All my machines retry regularly without the increasing delay times. I've started boinc.exe directly from the command line and via a command line shortcut with normal parameters. Both work here. These 2 commands have saved hours in delays when downloads stall multiple times. Now the stalled DLs usually complete in 7-10 minutes. Edit: Try using 60 seconds as previously suggested for both parameters. I noticed erratic behavior when trying settings of less than 60 seconds. |
|
Send message Joined: 30 Apr 13 Posts: 109 Credit: 3,977,737,860 RAC: 575,436 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
It looks like the --pers command runs the show. I don't see any evidence that the cc_config entry http_transfer_timeout actually does anything - at least not by itself. Do the two work together? The logs only show one delay: e.g. "backing off 00:00:50" which is now thankfully reduced by --pers. If you watch a download live in BOINC Manager there's an unlogged 5 minute delay (under ELAPSED in the picture below) with every hang BEFORE the backoff delay starts. So if one file hangs 3 times, that's another 15 minutes lost. This grab was made just after the 60 second retry began shown under STATUS. Now, it's great that the backoff delay is reduced to one minute, but I was hoping the 5 minute delay would also be reduced. Oh well, these are only workarounds to the bigger network issue. New Question: Is there a version of the --pers parameter that can be entered in the global_prefs.xml or global_prefs_override.xml file? e.g. <pers_retry_delay_max>60</pers_retry_delay_max> ?? I ask because you mentioned you have a shortcut in your startup folder. That's CURRENT USER or ALL USERS startup folder, right? I have Windows 10 and my BOINC starts somewhat elaborately (8 line items) in the SYSTEM startup (which has priority before the other startup folders), and I haven't figured out how to make any command line parameter entries there. So, for now, after a reboot, I manually shut down the client and restart via a shortcut which has the --pers command. Thanks. |
BeyondSend message Joined: 23 Nov 08 Posts: 1112 Credit: 6,162,416,256 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
New Question: No, if there was we'd be using it. I ask because you mentioned you have a shortcut in your startup folder. That's CURRENT USER or ALL USERS startup folder, right? I have Windows 10 and my BOINC starts somewhat elaborately (8 line items) in the SYSTEM startup (which has priority before the other startup folders), and I haven't figured out how to make any command line parameter entries there. So, for now, after a reboot, I manually shut down the client and restart via a shortcut which has the --pers command. Set BOINC manager not to start automatically. I simply make a shortcut on the desktop and copy it to the startup folder. I'm using Win7 and won't be using W10 if I can help it. I get no 5 minute delay. If you want to use BOINC Manager start it later. I find BoincTasks to be much more useful. Most of this has been covered already in these threads: https://www.gpugrid.net/forum_thread.php?id=4399 https://www.gpugrid.net/forum_thread.php?id=4373 |
skgivenSend message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Is this still an issue after the new server was installed (assuming it was; didn't notice anything)? FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help |
|
Send message Joined: 23 Dec 09 Posts: 189 Credit: 4,813,881,008 RAC: 42,053 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Yes, it persists. |
|
Send message Joined: 5 Jan 09 Posts: 670 Credit: 2,498,095,550 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Is this still an issue after the new server was installed (assuming it was; didn't notice anything)? Don't think they have done server upgrade yet. |
|
Send message Joined: 11 Jul 09 Posts: 1639 Credit: 10,159,968,649 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Is this still an issue after the new server was installed (assuming it was; didn't notice anything)? Test posted at exactly 18:20 Edit: post registered at 18:15:20. Either they've configured the new server with the same clock offset as the old one, or it's still the old one with the bad clock setting. |
BeyondSend message Joined: 23 Nov 08 Posts: 1112 Credit: 6,162,416,256 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Today the stalled downloads are worse than ever :-( |
©2026 Universitat Pompeu Fabra