I have a vb4 system with 5M posts and 300k threads. Running on a beefy enough system that frontend VM unix load is 1 with 8 VCPU, and db VM unix load is 0.2 with 4 VCPU.
I ama little concerned about a few things:
So... could anyone give me general expectation of how long this could take to import? Hours? Days? Weeks? Fugget-about-it? Do I start taking prozac now?
What can I do to help my odds?
Looking forward to this migration though, XF 2 looks awesome both front and backend especially when compared to VB4.
I am
- iSCSI SAN has 10+ spindles so sequential bursty access is fast, but random access and throughput sucks, limited by GbE
- The live VB system is super performant, but only because I allocate enough RAM for mysql to cache the entire database into memory, so SAN is irrelevant for read (forum is mostly read access, 10:1 guest to member ratio).
- I'm afraid that 5M writes to SAN will take forever. Assuming it took 10ms per post (am I dreaming in colors?) it's still 12 hours!
- I have 60 GB of attachments, because reason, will this ruin my life?
- It took 1 hour to import my 130k users, I hope the user import procedure is more complex than that of posts, or I am in trouble
So... could anyone give me general expectation of how long this could take to import? Hours? Days? Weeks? Fugget-about-it? Do I start taking prozac now?
What can I do to help my odds?
- Yes I read the comment about co-locating DB on the same machine as XF for import purposes. But I can't really do that, the DB server does DB and the web server does web. But of course I can do that, if I temporarily set up a mysql replica on the web node. Setting up mysql/innodb replicas is not in my top 10 things I enjoy doing, so, would this be worth the pain?
- What if I allocate crazy ram to my VM, do the import to a DB backed by ramdisk, then when completed replicate it to the DB node?
Looking forward to this migration though, XF 2 looks awesome both front and backend especially when compared to VB4.