Stderr output? ..server request to delete file...
9 Mar 2011 12:10:52 UTC
Topic 195697
(moderation:
Not sure why but one of my machines has deleted all of the files by 'server request to delete file' - this was probably near or at the end of the crunching.
Not too concerned however the einstien link of 'My Computers' for that pc still shows the task as active AND a new one which is currently running now BUT the manager on that machine is only showing the new task.
To make matters worse, this WU was first sent out to 2 pc's back on the 11th of feb - both of which timed out. A third machine picked it up on the 1st and completed already. My machine picked it up on the 3rd and now we are here.
Will my pc redownload that task or is it in never never land? If I don't snatch it back quick it will surely time out against that machine (its a slow one). I ALWAYS complete my tasks, its my mission.
I've included links - not sure how these will post.
task id
http://einsteinathome.org/task/221136140
work unit
http://einsteinathome.org/workunit/92940360
computer 1d
http://einsteinathome.org/host/3704209
'Old age and treachery will overcome youth and skill'
Phil
Stderr output? ..server request to delete file...
)
Did you check that the files are deleted? "Got server request to delete file ..." isn't acted upon before all tasks needing the data files are finished.
Gruß,
Gundolf
[edit]Don't confuse the data files (e.g. h1_1492.75_S5R4) with the tasks (e.g. h1_1492.75_S5R4__1218_S5GC1HFa_0)[/edit]
Computer sind nicht alles im Leben. (Kleiner Scherz)
Gundolf I have checked
)
Gundolf
I have checked more clearly -
I believe you are correct about the deleting files part, however the main thing that I noticed to begin with still remains. The manager on the pc thinks it has sent the file uploaded successfully (says 'finished upload of ....'), the einstien link to the My Computers shows that particular WU as not reported yet. In the messages there is no reference of that data being cleaned out yet, so is there a way to force another upload of that file.
I've got eight pc's, 7 at home and 1 across the river, plus I help admin 2 others all running einstein. I have never seen this before.
UPDATE (literally)
Well there was a new one - I decided to hit update (again) and einstien finally recognized the WU as reported(13 1/2 hours later), the Stderr ouput error on the WU link went away, nothing changed good or bad on the manager, and now I can get back to the rest of the farm. It's turning into a job, and I already got one. But I continue to have the urge to 'just add 1 more pc'. So I will!
Thanks for your support
Phil
'Old age and treachery will overcome youth and skill'
RE: ... however the main
)
This is simply 'the way things work' and it happens on every host, every time work is completed. Returning of completed results is (by design) a two stage process. Stage 1 (the uploading part) happens immediately and has low impact on the server because no database access is needed. The data is simply inserted into an upload area on the server. Stage 2 (the reporting part) will happen at some later time at BOINC's convenience (unless you force it prematurely with 'update'). The reason for this delay is to allow multiple results to accumulate in the upload area and then all be reported with a single (expensive) database access, which will most likely occur in conjunction with a new work request. Not having separate accesses for each result and each work request helps to lower the load on the server. When you force the issue with 'update', you defeat this strategy. No big deal really, unless many thousands of other participants were also doing the same thing :-).
It happens all the time and is particularly beneficial (and noticeable) with fast multi-core hosts. You won't see it on your hosts because they are mainly older, slower, single core machines where BOINC will have given up and reported the result before a second one has had a chance to finish and join it :-).
I know the feeling well :-). At one point 'just one more' had taken me to over 200 in total. I had a crush on Tualatin P3s :-). These days I'm down to around 80 hosts, but I'll probably build a couple more shortly. The Phenom II 840 quad core CPUs are nice and cheap and AMD CPUs perform very well on GW tasks. I'm planning to add a CUDA capable card to each build so I can have GW tasks on the CPUs and BRP tasks on the GPU. It's really quite a waste to do BRP tasks on CPUs when even a budget card like a GT430 should crunch them many times faster than even a modern CPU. You ought to consider replacing your fleet of 'oldies' with a single modern host with a suitable GPU. What you would save in electricity would pay for the new machine quite quickly and you'd end up with a far higher science output as a bonus. Take a good look at how long your hosts take to do BRP tasks - it's frightening! :-).
Cheers,
Gary.
You may think the electricity
)
You may think the electricity is frightening but the fact of the matter is I spend less on electricity now than when my X-girlfriend lived here! LMAO
On a more serious note:
You say it happens all the time but I never seen it before, maybe I'm just paying more attention to my 'job' these days. Yes, I can't help throwing a free piece of junk at the project - by the way....
I see a lot of 8 cpu machine's newly created and failing to return 1 successful task.
Anyways I am really trying to get more informed currently on the GPU concept but I lack in the sense that I never really cared about gaming - which is where these critters came from. My current girlfriend's (love of my life in case she see's this) nephew knows more than I do about that and doesn't know crap about the rest of the system, he just games. It'll come together in time I suppose.
I am really happy with the lastest aquisiton which is a hyperthreaded 3.0, seems to be cranking along just fine. Girlfriend likes it too since I installed it at her house. I was skeptical that it would perform well since it is a 'virtual' dual core. I haven't even tweaked it yet - got new dsl working, installed boinc, gave her the 'keys'. Told her I would try to break it next weekend - boy did I get a wierd look.
Thanks for your input very much - I think I just met someone crazier than me - '80 pc's'. (I know there are more - seen the stats)
Exactly how do you manage them efficiently? And that is a serious question I need answered. I find myself every morning and evening checking on the flock - and it takes time, sometimes a lot. Even if I upgraded I would still have multiples.
Phil
'Old age and treachery will overcome youth and skill'
RE: You may think the
)
I will give my own thinking here but with only 13 currently running pc's it is what it is...over time you will find that most Boinc only machines do not need daily checking. Yes sometimes they will need something tweaked but I personally use UltraVNC, it is free, and I just sit at one pc and log into all the rest of them from it. I usually check mine once a day but normally find them running just fine with no inputs required from me. I do manually do the Windows updates though, when they come thru I do them. I have it set to notify me but not download or install them without my input.
Also you can get a pci gpu that will increase your workunit output quite a bit meaning it will work on older pc's too! You can still find Nvidia model 9500 pci cards for sale on the net. For your newer pc's though a pci-e, also called pci-express, card is the latest and greatest and will pump out A TON of credits in a short amount of time! Pci and pci-e is the way it connects to the motherboard not a particular model of card. You can spend anywhere from waaay under 1oo bucks to well over 400 bucks on one card, and you are correct the better for games the better for Boinc. BUT you CANNOT game and Boinc at the same time on the same card! But you really can't game 24/7 so there is always time for Boincing!!