| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 19
|
|
| Author |
|
|
AlmightyMoe
Cruncher Joined: Nov 2, 2016 Post Count: 35 Status: Offline Project Badges:
|
Hi everyone,
----------------------------------------I recently experienced a hard crash on one of my devices, and as I was replacing storage to bring it back up again, I realized as it's the same device - don't want it to be listed as a separate listed from a tracking standpoint as it's only the storage that's changed. Anyone had that experience before? I generally don't go the extra mile to backup crunching machines, but perhaps there's a way to store it's footprint / computer ID in case I have a similar failure again? Any case, raised a case already to add my new computer ID, so the ask is not that. Wondering if there's something I can do on the client side so that I don't need to make IT support requests. Also I may be misremembering, but I swore that when IBM hosted the site, I may have had a similar issue but by naming the computer in the OS the same, it just worked. Maybe I'm imagining? Don't recall having to hunt for computer IDs before then. Anyway appreciate any perspectives. Thanks, ![]() |
||
|
|
hchc
Veteran Cruncher USA Joined: Aug 15, 2006 Post Count: 865 Status: Offline Project Badges:
|
I can't find the URL to the WCG article that describes their custom matching process, but it does go off the hostname in the operating system and some other factors. It's not always perfect. I have a duplicate Device ID and because WCG deviated from vanilla BOINC, there's nothing I can currently do to merge the two's statistics into one device/host ID. Frustrating.
----------------------------------------Try installing BOINC on the new storage device, but before attaching it to WCG, restore the data directory. On Windows, that's usually C:\ProgramData\BOINC
|
||
|
|
AlmightyMoe
Cruncher Joined: Nov 2, 2016 Post Count: 35 Status: Offline Project Badges:
|
Yup, I was more thinking of failure (as in my current experience) than transfer. I'm sure there are some instructions on how to do that, but was thinking more how to mitigate when a crash strikes.
----------------------------------------I did find something on the boinc site itself, but as with most anything there, unclear if that was how it worked with how it worked with IBM, or it's how it works today. Namely the same OS (I upgraded), username, domain name, ip address (that depends on ISP), processor vendor and memory. Anyway sounds like from your experience, they are going to look like 2 different devices and I'm going to have to live with it. ![]() |
||
|
|
BobbyB
Veteran Cruncher Canada Joined: Apr 25, 2020 Post Count: 638 Status: Offline Project Badges:
|
My take is that the domain name and host_cpid are what does it. By domain name I mean the name of the machine. It is the name you gave it when installing Windows. For Windows it is the name in the system/about settings or host name using ipconfig /all on a command prompt.
----------------------------------------My Linux machines use the names I gave them when I installed Unbutu. The host_cpid is in client_state.xml as is the domain name So my guess is that if you save C:\ProgramData\BOINC and give the new machine the same computer name, you are in business. The rest OS, username, ip address, processor vendor and memory are irrelevant. How I would do it is install Windows using the same name, download Boinc, disconnect the Internet, install Boinc, then fix client_state.xml with the host_cpid of the old machine and check if the domain name is the same as the old one.Then I would re-boot in case some setting was "remembered" in those modules which start when the machine is booted. I've never tried this. I know for fact that I've moved the HD of a Linux machine from a lesser machine to a more powerful one and it just continued to work. Of course I emptied the queues first. Linux must have done some calibrating on the first boot. I went from an Intel to an AMD. Done this more than once. [Edit 4 times, last edit by BobbyB at Oct 30, 2023 11:40:39 PM] |
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 2173 Status: Offline Project Badges:
|
My take is that the domain name and hostid are what does it. By domain name I mean the name of the machine. It is the name you gave it when installing Windows. For Windows it is the name in the system/about settings or host name using ipconfig /all on a command prompt. What you are referring to is NOT a "domain name' but a "host name"My Linux machines use the names I gave them when I installed Unbutu. The hostid is in client_state.xml as is the domain name So my guess is that if you save C:\ProgramData\BOINC and give the new machine the same computer name, you are in business. Well, apparently not.The rest OS, username, ip address, processor vendor and memory are irrelevant. I have one machine currently crunching WCG, my (somewhat older) MacBook Pro laptop. This has a host name of MacPhoenix (as it is resurrect from the ashes several years ago) that gets a dynamic IP (private) via DHCP on the LAN network here in the office where I am usually work each day, for the last 6 months or so. However, in the Results page, it does show up as "ojmr-4a.ojmr.corp", which is a host name of a machine gone to the eWaste/electronic recycling about 6 or 7 years ago, back then a PC workstation running Windows XP. There is only one thing that those two hosts could possibly have in common at this point, and that is the local LAN IP address from those several years ago, as even the public IP address of the office has changed since then... Doesn't really bother me, as it still refers now to the right machine, even showing the right version of macOS (instead of Windows XP) in the Results. But it clearly shows that there must be something else but host name and host ID that is being used to identify a host in the databases... Ralf |
||
|
|
AlmightyMoe
Cruncher Joined: Nov 2, 2016 Post Count: 35 Status: Offline Project Badges:
|
My Linux machines use the names I gave them when I installed Unbutu. The host_cpid is in client_state.xml as is the domain name So my guess is that if you save C:\ProgramData\BOINC and give the new machine the same computer name, you are in business. The rest OS, username, ip address, processor vendor and memory are irrelevant. How I would do it is install Windows using the same name, download Boinc, disconnect the Internet, install Boinc, then fix client_state.xml with the host_cpid of the old machine and check if the domain name is the same as the old one.Then I would re-boot in case some setting was "remembered" in those modules which start when the machine is booted. Honestly I've never upgraded a device without changing it's domain name or machine name usually, however I get what you are saying, and it makes sense. Good to know it works. So as I didn't save the client_state.xml given the crash (it's a long story, I could have saved it after doing a part magic fix, and as I thought it worked, I let it run until it hard crashed again, and then it was permanently lost), the only thing I could do was name it the same with the linux install. Short story - it's working now. Pretty sure the machine ID did change, and along with the upgrade to the OS using a later distro build, everything else was the same. Whatever it did to determine if the machines were the identical, it worked. I guess I will save that xml file for future safekeeping, and given this experience whatever automated backend work that was done or if indeed it was the tech team who rectified it, I'm happy. Still can't say for certain what it was though, so if I ever have this happen again, I'm hopeful I'll have a similar outcome. ![]() |
||
|
|
BobbyB
Veteran Cruncher Canada Joined: Apr 25, 2020 Post Count: 638 Status: Offline Project Badges:
|
@TPCBF Just to be clear, I was refering to what it says in client_state.xml. It is misleading though and I should have written host_cpid instead of hostid. I edited it a little later.What you are referring to is NOT a "domain name' but a "host name" <domain_name>xubuntu2</domain_name> This it's WCG name in the results page and also the name I used when I installed Ubuntu. I have trouble accepting that your macbook just happened to get the same private IP as an XP machine from 6-7 years ago and this is what solved it. A DHCP assigned LAN IP is NOT something I would use to id a machine. There is a WCG device ID associated with each of your machines. You see it in the logs when it boots or click properties for WCG in BoincMgr. In other threads where people complain about missing devices this number is what they give to tech. https://www.worldcommunitygrid.org/forums/wcg...d_thread,44607_offset,470 I don't know how their system decides that it is a new machine when it connects at boot but it can't be a LAN IP. What is unique in client_state.xml is <domain_name> and <host_cpid>. But this does not explain your situation. @AlmightyMoe OOPs that was Linux and I thought it was Windows. I've done the HD move 3-4 times but never for a crash. Guess I will save the host_cpid.[Edit 2 times, last edit by BobbyB at Nov 2, 2023 5:34:26 PM] |
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 2173 Status: Offline Project Badges:
|
@TPCBF Just to be clear, I was refering to what it says in client_state.xml. It is misleading though and I should have written host_cpid instead of hostid. I edited it a little later.What you are referring to is NOT a "domain name' but a "host name" <domain_name>xubuntu2</domain_name> This it's WCG name in the results page and also the name I used when I installed Ubuntu. I have trouble accepting that your macbook just happened to get the same private IP as an XP machine from 6-7 years ago and this is what solved it. Well, nothing is "solved" in that case.A DHCP assigned LAN IP is NOT something I would use to id a machine. Well, Bobby, I do networking for about 40 years now. I do know how to set up computers on IP networks.Yes, if it is a desktop computer that doesn't move, I will prefer to give it a static local IP. As I commonly in such case also need to set up firewall rules for remote access, for example. However, the MacBook is, as you might know, a laptop, and while it currently sits quietly in a shelf and crunches along, it is not intended to be permanetely be in this location. At times, I take this laptop with me, for example when giving talks about cross-platform programming or need to do macOS related stuff at a client, and thus by default, it is set up to get an IP via DHCP. On both wired and wireless connections. That old workstation host name as I mentioned is from a long gone machine, dual core P4 IIRC. And back then, the office was much larger as it is now, with many more computers, the DHCP range has likely changed (it is bigger now, as we have less static IP workstations, as back then). And that had back then a static IP and was for a while also running BOINC/WCG it's it's "spare time". There is a WCG device ID associated with each of your machines. You see it in the logs when it boots or click properties for WCG in BoincMgr. I not only know how networking works, I also know quite well how BOINC works, I am using it for pretty much two decades (almost 13 years now for WCG). But...Please download the CSV file with current WUs in the "Results" list, and you will see that there is NO device ID in that list (and that is what I am talking about), it just has a device name. Here's the header (and the first data record) of that CSV file to save you the download: metadata This is the data record of a WU just send to that device, hence some of those data fileds have default (non-)values, like appVersionNumber, credit, etc...limit,pageNumber,recordsAvailable 0,1,10434 items appVersionNumber,claimedCredit,cpuHours,deviceName,dueTime,elapsedTime,grantedCredit,os,osVersion,owned,resultId,resultName,returnedTime,sentTime,st atus,statusName,workunitId 0,0.0,0.00000,OJMRDATA,"2023-11-09T14:30:14+0000",0.00000,0.0,"Microsoft Windows Server 2016","Standard x64 Edition, (10.00.17763.00)",,712543850,MCM1_0207404_9000_1,,"2023-11-03T14:30:14+0000",0,"In Progress",415933255 There is no deviceID, and as device IDs are AFAIK assigned increasingly, with old numbers not being reused (for obvious reasons), so the only common nominator between that old XP workstation and that MacBook laptop (which came in my possession AFTER that XP machine was already decommissioned!) can only be the local IP number... Ralf |
||
|
|
katoda
Senior Cruncher Poland Joined: Apr 28, 2007 Post Count: 172 Status: Offline Project Badges:
|
Do not forget about very important parameter - rpc_seqno. You can try to make your new machine as close to the old one as possible in terms of IP, name, OS etc, but if your "new" machine will have incorrect rpc_seqno set in BOINC config file, it will be ALWAYS registered as a new device.
----------------------------------------I shortly described my findings here https://www.worldcommunitygrid.org/forums/wcg...ead,45386_offset,0#686763 In the past I did various stuff, like moving BOINC profiles between different PCs, with different H/W configs (CPU included), heck, once I moved Windows profile to a Linux machine and always the new machine was correctly recognised as an existing device, so WCG system is not that picky as some could suspect. In the specific situation discussed here, I would try to backup BOINC profile as frequently as possible, in the same time trying to estimate rpc_seqno increase between backups. In case of crash, just restore the most recent backup of the profile, set rpc_seqno to the number which is certainly bigger than last possible real value of this parameter (eg. mean increase between backups is 1000, so in case of restoration of the profile from the backup increase it by 1500 or even 2000) and that's it, should work. ![]() |
||
|
|
BobbyB
Veteran Cruncher Canada Joined: Apr 25, 2020 Post Count: 638 Status: Offline Project Badges:
|
Well, Bobby, I do networking for about 40 years now. I do know how to set up computers on IP networks Ditto here.I use static IP also for management purposes. I use the MAC for this. I still maintain that the IP is not something I would use to ID a machine. Most people don't do static. download the CSV file with current WUs in the "Results" list Did that and it is like yours.Download the device list. Mine shows this: (the last 2 are dead) metadata limit,pageNumber,recordsAvailable 0,1,7 items id,lastResult,name,os,points,registrationDate,results,runtimeInSeconds,type,venue 6925894,"2023-11-10T14:04:55+0000",xubuntu2,"Linux Ubuntu",121915072,"2020-07-08T01:06:41+0000",198332,1284026738,B,work 6778353,"2023-11-10T14:04:15+0000",ubuntu,"Linux Ubuntu",26729649,"2020-04-25T19:22:59+0000",44444,520451273,B,home 6782051,"2023-11-10T13:47:59+0000",x-ubuntu,"Linux Ubuntu",40839472,"2020-04-27T21:04:08+0000",71426,507445623,B,home 7687670,"2023-11-10T13:38:00+0000",HP8460-PC,"Microsoft Windows 10",9851448,"2021-02-02T18:16:20+0000",16105,208267063,B,Default 8689953,"2023-11-10T13:20:45+0000",HP-PC,"Microsoft Windows 10",2558314,"2023-01-18T01:01:35+0000",4189,70109939,B,Default 6778504,"2021-03-20T19:59:02+0000",BOBT400-PC,"Microsoft Windows 10",2210931,"2020-04-25T21:20:18+0000",3852,48437242,B,home 7392510,,Bob-R6PC,"Microsoft Windows 10",0,"2020-09-29T18:18:28+0000",0,0,B, The explanation given by katoda makes sense. I was in on that thread. My experiments were using the same hard drive. Never did a crash. This seems like the missing link to how Boinc identifies the devices. On a new install this rpc_seqno would not exist so it would see it as a new device. I'm guessing that if you did a new install and wanted it to look like the same device the info in client_state.xml would do it. Still does not explain how it saw the Macbook as an XP machine. A fluke somehow. I'm not going to experiment with moving profiles around like katoda did just to see what happens. [Edit 2 times, last edit by BobbyB at Nov 10, 2023 4:58:05 PM] |
||
|
|
|