Recommendations when Converter 3.0.1 start FAST but starts getting SLOW

[benny_hauk]

I've got a physical machine I'm trying to do a hot P2V on during production hours. Understandably the conversion process was real fast for the first 30% and now it's dog-slow. I'm fine with that and totally expect it. Every now and then it speeds up a bit then slows down. No problem.... I think.

My question is whether this is a good strategy or not. Is there something about the snapshotting process that has problems if I let a conversion run for, say, 24 hours? In the V2V world, snapshotting creates redo logs which, given enough time, can cause serious problems (like when I forgot about one and let it grow to over 20GB one time). I am resizing (to smaller drives on the new VM) so I suppose it's doing a file-level copy and not a block-level. If some kind of "redo log" type operation is happening on the physical box with the VMWare Converter Service then I could see how at some point performance would go down the drain or I would hit some resource limit - if that's the case then I'd just prefer to wait until afterhours and cold migrate it. I don't know how it "snapshots" a physical server, so I guess that is my question. Is the physical disk in some sort of strange state while the snapshotting is occurring? If it's just copying files and it doesn't actually take a single snapshot of the disk, then I don't get a consistent state of the files (some may be more recent than others, etc) and this would be a bad thing. The server being P2V'd is Windows 2000 so snapshotting at the OS level isn't supported I don't think as it would be with Windows 2003.

Can anyone provide some explanation? A short answer would simply answer the question of whether it's best to P2V when you anticipate the process taking the quickest amount of time possible or whether it is okay to just let it run slow and take hours to complete. I'm okay if there's some overhead during the P2V process during the day but if the overhead gets worse with time then that would be a major issue.

(I'm not really asking for best practices here, that's a no brainer. Just needing to know if I'm going to shot myself later on for letting it run for a long period of time)

Thanks!

[esiebert7625 · Accepted solution]

What type of server is it? If it is fairly static content on the drive then it is not a problem to let it run a long time. If it's a database/Exchange server then you might be a little more risky, especially if they are running during the conversion. Typically it will slow down when doing a file by file copy when it hits alot of small files. Mine slowed to a crawl because I had one directory that had 100,000+ 2k smtp badmail files. If it's possible you might want to clean up any data you can beforehand. The below post describes the snapshot process that Converter does. Bottom line, I don't think you'll have a problem letting it run a long time. I have done dozens of conversions some taking up to 15 hours and have not had a problem afterwards.

http://www.vmware.com/community/thread.jspa?threadID=87540&messageID=660907#660907

[theanykey]

if you resize the disk then it will do a file-by-file copy

if you maintain the disk size then it will go faster using a block-by-block copy

you may want to defrag the pc 1st if you use a file-by-file copy