My browser gets nothing from the ABP address, and the Firefox status area on the bottom bar simply says "Waiting for atlas1.atlas.aei.uni-hannover.de...
As something about the situation seems to be suppressing most download requests (my queues have shrunk a lot)...
Over something like the last couple of hours, the success rate of upload attempts from my hosts has been materially better than for the past several days. So much better that the number of uploads awaiting retry has shrunk.
It appears that, at least for my hosts (running 5.10.20 BOINC), that they resumed requesting work when the number of failed uploads awaiting retry went below a threshold, possibly of about ten tasks.
For about the last 15 minutes on my hosts, it appears that nearly all upload retries have succeeded, though with appreciable delays.
Also, the considerable amount of download caused by resumed work request has been succeeding at a considerably higher fraction of attempts than a few hours ago.
While transfer rates and response times remain slower than usual, I'd guess that may be a side effect of the very high pent up demand manifesting as retry requests. May we thank someone for fixing something?
This is weird, I only see a single(!) try in access.log and nothing at all in error.log - thus I suppose it's never really hit the apache here in Hannover.
From the time stamps, I'd guess that only the last (successful) packets came through to the apache logs. [edit]My system clock gets updated twice a day by a time server, so should be pretty accurate.[/edit]
There have been more download attempts for that file, but I only copied one of each kind to keep the post short.
That sounds to me as if something "in front of" the web server kept the connections from getting through, as Richard already suggested.
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
Is it just that it is taking a while for all the work to make it down?
On my hosts, work request was disabled for a couple of days, apparently by a rule invoked by more than about ten upload retries pending. So when things got better yesterday, the work requests made once my pendings dropped below ten were huge. If I recall, each ABP task required the download of four 2 Megabyte files. A huge request could get many such files, which take a while to transfer. I had a couple of systems which spent more than ten minutes in nonstop (successful) download.
Not the problem of main
)
Not the problem of main concern, but I'll mention just in case it might provide a clue.
The main Einstein status page has a link to ABP overall search progress
My browser gets nothing from the ABP address, and the Firefox status area on the bottom bar simply says "Waiting for atlas1.atlas.aei.uni-hannover.de...
RE: As something about the
)
Over something like the last couple of hours, the success rate of upload attempts from my hosts has been materially better than for the past several days. So much better that the number of uploads awaiting retry has shrunk.
It appears that, at least for my hosts (running 5.10.20 BOINC), that they resumed requesting work when the number of failed uploads awaiting retry went below a threshold, possibly of about ten tasks.
For about the last 15 minutes on my hosts, it appears that nearly all upload retries have succeeded, though with appreciable delays.
Also, the considerable amount of download caused by resumed work request has been succeeding at a considerably higher fraction of attempts than a few hours ago.
While transfer rates and response times remain slower than usual, I'd guess that may be a side effect of the very high pent up demand manifesting as retry requests. May we thank someone for fixing something?
RE: This is weird, I only
)
From the time stamps, I'd guess that only the last (successful) packets came through to the apache logs. [edit]My system clock gets updated twice a day by a time server, so should be pretty accurate.[/edit]
There have been more download attempts for that file, but I only copied one of each kind to keep the post short.
That sounds to me as if something "in front of" the web server kept the connections from getting through, as Richard already suggested.
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
The dam seems to have
)
The dam seems to have broken...no waiting UL or DL at the moment...
Yep, I too got a new ABP this
)
Yep, I too got a new ABP this morning - without any retry.
Did the problem correct itself or did someone find the screw to turn?
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
RE: Yep, I too got a new
)
I've given the apache more threads to serve requests, the load is now higher but if the errors are going down that's good :)
Now got a real wadge of work
)
Now got a real wadge of work I cannot download.
Help
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
RE: Now got a real wadge of
)
Is it just that it is taking a while for all the work to make it down?
On my hosts, work request was disabled for a couple of days, apparently by a rule invoked by more than about ten upload retries pending. So when things got better yesterday, the work requests made once my pendings dropped below ten were huge. If I recall, each ABP task required the download of four 2 Megabyte files. A huge request could get many such files, which take a while to transfer. I had a couple of systems which spent more than ten minutes in nonstop (successful) download.
What status do you see under Transfers?
Slowly downloading although
)
Slowly downloading although only one file at a time rather than two.
Annoyingly it gave up on the exe file and I've got plenty of WUs downloaded with plenty more to go but no app to run them :)
RE: Slowly downloading
)
ABP data files are served from one machine only, which is under administration of the ATLAS Cluster admins.
The application files, however, are served from our download mirror network, which is under heavy load right now (see front page news).
BM
BM