Download issues with Gravity wave work
15 Sep 2020 11:21:53 UTC
Topic 223499
(moderation:
I'm having issues downloading the skygrid and data files for Gravity wave work. I get allocated the files to download and then for each file it times out after 2 minutes and cycles through the files. It repeats this for each file a second time before it eventually gets them on the 3rd attempt.
I turned on http debug and apart from getting 6000 lines of output it mentions GET http://einstein-dl.phys.uwm.edu/download/24b/h1_0145.35_O2C02Cl5In0.HsGB (the filenames vary). If I type the address into a browser I get a "URL cannot be found message". I have a recollection that some of the servers were moved from uwm to Germany (mentioned in a BOINC conference call). I'm wondering if its trying the uwm address first and eventually timing out before trying the servers in Germany.
Looking in client_state I see there are 3 download URL's listed for a recent skygrid data file as:
http://einstein-dl.phys.uwm.edu
http://einstein-dl4.cgca.uwm.edu
I'm guessing the first two fail and syr.edu is found.
Update. I tried a ping to
)
Update. I tried a ping to each of the addresses above. While the addresses resolve the first two uwm.edu addresses time out. The third one syr.edu resolves and responds to the ping.
BOINC blog
Did you ever resolve this
)
Did you ever resolve this issue? I just started downloading GW tasks again and I have 30 or so stuck file downloads that are not showing any progress for the past couple of hours. Restarting BOINC or the computer changes nothing. I assume the files are trying to be downloaded from the two servers you said weren't responding.
I have seen a similar
)
I have seen a similar condition a few times in the past week. A whole "stack" of downloads are queued up in the "Transfers" list. Each goes to "download active" for 34 seconds then reverts to "download pending" having made zero percent progress. The process goes through the entire list and cycles through again. It does resolve itself after some undetermined time; as when I might check on progress a few hours later all transfers have completed. The next time I see this I'll try to captue some diagnostic information - for instance, capture the outgoing packets for destination address and match up with replies, if any.
Well I gave up after 8 hours
)
Well I gave up after 8 hours of no progress. I just had the tasks in Active download state and pending and no progress. So I aborted them and turned GW back off. This was the same state on both my hosts.
Keith Myers wrote: Did you
)
No. fortunately I run a small cache (0.2 days) due to Rosetta so I only get enough tasks for each thread in the machine. I switched back to Rosetta.
It has to retry each task at least twice before moving through the entire list before it gives up with the first download server. It theN repeats for the second server. So two mins per task x 30 tasks x 2 download servers is 120 mins just to start the download.
BOINC blog
Well I waited much longer
)
Well I waited much longer than 2 hours on both machines and no file ever started downloading.
I wonder what has changed with the servers. The last time I ran GW tasks, they uploaded and downloaded fine.
Anybody ever send a message to the admins?
Just started a new batch of
)
Just started a new batch of GW Gpu tasks and they were downloading fine for me.
bigendian81 wrote: Just
)
New batch meaning just more work that you have already downloaded the seed files for?
Or new batch meaning you had stepped away for awhile and now just starting back up and have to download a new batch of seed files?
That was the situation I ran into. I had to download all the precursor files for the current series they are sending out. Those were the files that would not download. I had none of the files in the directory needed for the new work.
No problem with downloading
)
No problem with downloading on a fresh install of boinc on this machine. There would not have been anything cached on this one.
OK, thanks for the post. So
)
OK, thanks for the post. So then just bad luck that the tasks I tried to get had bad pointers to where they were stored across the 3 different Einstein download servers.