| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: What causes a change of ID and how do we stop it? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 24
|
|
| Author |
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
I was a surprised myself that the 6 items that WCG tests to find the old registration did not kick in. Originally i read that the out of sequence counter was used to merely erase the work, but came out for it not to be. Now i understand.... if it is a misidentification and really 2 running around with same id, because there was a none-zero state out of sequence contact, it would become a mess.
----------------------------------------So, a backup needs to be made and no contact should be established until the restore on the new device is complete (The FAQ actually tells to do this off-line). How CPDN hack it together has always mystified me (never tried to understand either) when they go on about model restores and resumes and frankly, not interested in their solution. The jobs are too small here to worry about it. Even on a quad statistically 49% tops average completion is lost per job in progress. The rest of the queued work would just be resubmitted, sooner or later. Anyway, wanna hack sequence numbers to save a device registration.... not going to explain that in any FAQ here. 2 Euro cents
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
(Sekerob)So, a backup needs to be made and no contact should be established until the restore on the new device is complete. In the case of a hard disk failure this is, of course, impossible. So, what you are saying is that BOINC on WCG totally fails to recover nicely from a system crash. (Keck_Komputers)It is also worth mentioning in that it is normally not possible to merge the host records when the sequence numbers get out of order. This prevents the same host record from being used for 2 computers when the BOINC directory is copied over to install BOINC on a new host. There are heaps of sensible ways of detecting that. The host name, IP address, etc, will have changed if it's a different machine. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello Kremmen,
One of the most puzzling problems that sometimes strikes members is when the server thinks that 2 different machines are the same. It can take days to figure out why the member is suddenly getting errors because the server assumes that commands that the member is giving on one machine should apply to the work units on the other. Naturally, computers are always moving to new IP addresses and members expect the server to correctly keep track of their laptop. It looks simple, but since CPU manufacturers do not manufacture each CPU with an unique ID, it is actually very complex. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Naturally, computers are always moving to new IP addresses and members expect the server to correctly keep track of their laptop. If the IP address has changed, that might be a reason to question whether the computer itself has changed. However, if the IP address (and host name and everything else) has not changed, it's a good reason to assume that it is the same machine. It would make sense to do the current host matching regardless of the RPC sequence number. The risk that a cloned machine will be almost identical to one it was cloned from and happen to have exactly the same IP address is tiny, yet to combat that tiny risk, we have our backups rendered useless within a few hours and new host records generated when we restore from backup. |
||
|
|
|