Is there a way to gracefully quiesce the Rosetta application? You know, basically say finish what yoiu are working on but don't start anything new off the work queue?

Because restart is not always 100% reliable (especially is you are putting the hardware on the bench to dink with it) I prefer not to "shut down connected client" or use the line command to do the same thing.

The "No New Tasks" button just says not to pull any new work off the server (great for phasing out of SETI) but work on the queue still continues to process which means you really have to plan ahead to idle up a system.

Any thoughts?

Hi Chris.

You could just hi-light the tasks you don't want to run and suspend them.

To easy :)

The only other way i know of is not download to start with ;)

---

You could just hi-light the tasks you don't want to run and suspend them.

Hey - thank you so much for the response - but isn't suspending them the same as cutting a checkpoint and quitting the BOINC client?

every time I see you Chris, you have a higher RAC!

---

Hi Chris.

I thought you where talking about stopping new tasks from running.

Yes if you suspend a task it will go back to the last check point.

That's why i run 4hr runtime, less chance of losing a lot of time. :)

---

The Chilean said:

every time I see you Chris, you have a higher RAC!

The RAC is mostly the result of a medical condition which has kept me pretty much grounded for the past few months - I am lucky that for a large part, I can work out of the house but my recreational endeavors have pretty much been limited to a weekly trip to the local electronics store to snap up whatever it is they have on sale.

At this point I think I am done putting new systems together for a while but I do have some real opportunities to tune and optimize a few under performers. Thats the fun part anyway. And was the root reason for my question on "clean" shutdowns.

But you know, RAC is not that important - I envy you your stunning graphic. I suspect that there are more than a few of us Texans scratching our heads wondering where you guys came up with your version of the Lone Star.

Well done!

Is there a way to gracefully quiesce the Rosetta application? You know, basically say finish what yoiu are working on but don't start anything new off the work queue?

Because restart is not always 100% reliable (especially is you are putting the hardware on the bench to dink with it) I prefer not to "shut down connected client" or use the line command to do the same thing.

The "No New Tasks" button just says not to pull any new work off the server (great for phasing out of SETI) but work on the queue still continues to process which means you really have to plan ahead to idle up a system.

Any thoughts?

You could go in and set the project to NNT then hi-lite all those you have not started yet and abort them, that way you would finish the tasks you are currently running and not get any new ones. And by aborting and then reporting those tasks you will be putting them right back into the queue for someone else to crunch. If you have several projects on the same pc you could click suspend and it would not process that one any more, until you go back to normal.

I am unclear what you mean about restart not being 100% reliable. Before you modify your machine, it would be good to take a backup of everything... including your BOINC directories. That would assure you can restore.

Nothing you can do will force a checkpoint. If you flag the project for no new work, then suspend any tasks that have not yet begun processing, then the tasks in progress will run through to complete so you can make your mods, then resume the other tasks when you come back up and again allow new work.

---

Rosetta Moderator: Mod.Sense

I am unclear what you mean about restart not being 100% reliable.

Hey, I was not trying to disparage BOINC's checkpoint / restart facility at all (but I will make an attempt to if you would like ...)

It is just that restart does not seem to respond well when I am off doing things that I should not do. Which is understandable.

For example, taking a system down and replacing the old and moldy Athlon processor with more up-to-date Phenom II six-core processors. Same family, different micro-architecture.

I started out running Rosetta on a rag-tag collection of processors built from my rather large assortment of cast off parts I have acquired over the past few years. About the only things new were the power supplies and cases.

As time and money permit I have been upgrading components here and there to improve stability and performance. Since I stayed within the AMD family I have not yet had to reload a system and unlike Windows, Linux happily accommodates major hardware upgrades.

That is what I meant when I used the term "dink with the hardware"

However, I don't know how checkpoint / restart works - if it stores an array of values reloaded at restart time or if it stores actual memory snapshots. it just seems that when coming back from a hardware tweak restarted tasks seem to have their problems.

Hence my request on info about how to quiesce the system gracefully.

(sorry if I am a bit verbose - I am a government contractor so it is required of me)

Thanks

Previously, the fat guy said to the Chilean ...

I suspect that there are more than a few of us Texans scratching our heads wondering where you guys came up with your version of the Lone Star.

With a deep and humble bow I have to acknowledge that your "lone star" predates ours by 21 years. Who would have thunk it?

Chris

When you change the CPU (or OS) type, I believe BOINC detects this upon first startup and ends up getting a new host ID, and then the work you have on deck is no longer for the ID of the host that is currently running.... ultimate result would be any work that did run and attempt to report back would be rejected, because it was never sent to this new host ID.

My approach was attempting to avoid aborting any work, but it's not really a big deal either way. If you are changing CPUs, aborting unstarted tasks may be the cleanest approach.

---

Rosetta Moderator: Mod.Sense

When you change the CPU (or OS) type, I believe BOINC detects this upon first startup and ends up getting a new host ID, and then the work you have on deck is no longer for the ID of the host that is currently running.... ultimate result would be any work that did run and attempt to report back would be rejected, because it was never sent to this new host ID.

My approach was attempting to avoid aborting any work, but it's not really a big deal either way. If you are changing CPUs, aborting unstarted tasks may be the cleanest approach.

When I do the 'dink of the computer' at my house I try to run Boinc out of work first to avoid these kinds of problems. I often wipe the hard drives when I do the 'dinking' so it is not a 30 minute process for me. I have done what Chris has said and have run into the problems described by Mod Sense too! I now either set the project to no new tasks or abort any still around before starting the 'dinking'.

Everyone - thanks for the responses. The information is great.

However, just to set a few items straight.

1. Linux does not require me to reinstall the OS as long as I stay within the same architecture family - in my case 64 bit X86 - its upgrade and go. I won't profess to know anything about how Windows handles the same situation but suspect that it would complain in some manner as it might see it as a "new system" and a potential licensing issue.

1a. Work that has not yet been started survives the reboot and hardware changes without any issues.

2. When performing a hardware upgrade without a new OS install I DO NOT get a new Computer ID - in fact as soon as I do a "run benchmarks" and a "update" Rosetta's reports reflect the new configuration - no change in host name or ID.

3. Terminating and / or aborting work is really not an option for me - I know the work is resent to another host in due time but it seems as uncouth as pissing in a punch bowl. Pardon my French. Even when I grew frustrated with SETI over all their outages and lack of communication I let the queues "bleed down" and exited gracefully.

I am not thrilled about waiting out the work queue for a day or more but at this point that seems to be the only option I have.

Once again, thanks for the input.

Previously, the fat guy said to the Chilean ...

I suspect that there are more than a few of us Texans scratching our heads wondering where you guys came up with your version of the Lone Star.

With a deep and humble bow I have to acknowledge that your "lone star" predates ours by 21 years. Who would have thunk it?

Chris

It's alright, I'm pretty sure they took the idea from the US flag somehow when figuring out what to use as a flag. Just one star for being a unitary republic state and not a federal state. And two stripes, just so that the copy-paste of the flag weren't as obvious Haha :D

And regarding the thread, I always do the "No New Tasks" to flush out the queue when switching to Linux to never use Windows again (which lasts a few months until I change back). Does dinking the hardware cause problems with the current WUs under Linux?

---

Chris, from what I've seen when a task is aborted by a user, it is typically sent to another within about 10 minutes :) So, certainly not a behavior to encourage people to do every day, but no big deal for the few times in a year you are going to be making such changes. Your time is worth something too. So make your hardware changes when it is right for your schedule, regardless of the size of your backlog.

---

Rosetta Moderator: Mod.Sense

When you change the CPU (or OS) type, I believe BOINC detects this upon first startup and ends up getting a new host ID, and then the work you have on deck is no longer for the ID of the host that is currently running.... ultimate result would be any work that did run and attempt to report back would be rejected, because it was never sent to this new host ID.

You can upgrade with no problem under Windows as well.

I just had the chance to test this. I replaced the motherboard, the CPU and the RAM in one of my computers. I didn't have to reinstall Win 7, since it's the same chipset (basically the same board, just two generations later). Even the RAID array was still functional (I wasn't sure, so I let BOINC run out of work yesterday).

Anyway, the CPU-ID remained the same, just the CPU type and the amount of memory were updated.

cu Joe

---

When you change the CPU (or OS) type, I believe BOINC detects this upon first startup and ends up getting a new host ID, and then the work you have on deck is no longer for the ID of the host that is currently running.... ultimate result would be any work that did run and attempt to report back would be rejected, because it was never sent to this new host ID.

You can upgrade with no problem under Windows as well.

I just had the chance to test this. I replaced the motherboard, the CPU and the RAM in one of my computers. I didn't have to reinstall Win 7, since it's the same chipset (basically the same board, just two generations later). Even the RAID array was still functional (I wasn't sure, so I let BOINC run out of work yesterday).

Anyway, the CPU-ID remained the same, just the CPU type and the amount of memory were updated.

cu Joe

I upgraded my machine from Windows XP to Windows 7, and I now have a different ID. My RAC dropped dramatically, and the screen saver no longer works, so upgrading is not totally transparent. When I check my stats, it shows results from two different computers, even though I have the same hardware.

deesy

I upgraded my machine from Windows XP to Windows 7, and I now have a different ID. My RAC dropped dramatically, and the screen saver no longer works, so upgrading is not totally transparent. When I check my stats, it shows results from two different computers, even though I have the same hardware.

If you reinstall the OS, your computer will always get a new ID, that is for sure. The question was, wether the computer gets a new ID, when only the CPU is upgraded.

If your RAC dropped dramatically, there is probably something wrong with the installation. All drivers installed? I noticed a slight increase when I changed from XP (32 bit) to Win 7 (64 bit).
As well the benchmarks of your machine is almost the same with XP and 7, so are you sure, there is a drop in your RAC, or is it just slowly increasing to the old value (this is what I would expect).

As well you could merge your computers.

Can't help you with the screen saver, I don't even have it installed on my computers...

cu Joe

---

I believe the RAC drop is simply due to a new host ID starting at zero and having to work up a 2 week average of work completed.

---

Rosetta Moderator: Mod.Sense