The Community Forums

Interact with an entire community of cPanel & WHM users!
  1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Trying to get close to 0 downtime on acconut transfers

Discussion in 'General Discussion' started by dspillett, Jun 14, 2007.

  1. dspillett

    dspillett Active Member

    Joined:
    Oct 2, 2005
    Messages:
    26
    Likes Received:
    0
    Trophy Points:
    1
    I'm about to transfer a number of accounts to a new cPanel install on a different server. The source server has a few spare IP addresses at the moment, so I was wondering if this would work for the highish priority accounts...

    PART 1:
    1. Give the account a spare dedicated IP address
    2. Wait for a bit (two days if I've not played with TTL values) to allow of DNS propagation delays

    PART 2:
    3. Block the dedicated IP with iptables (to stop mail and so forth being updated during the transfer) and stop any its cron jobs
    4. Transfer the account to the new location and set to the new location's shared IP
    5. Remove the iptables block on the dedicated IP, and replace with DNAT rules to forward packets to the new server
    6. Update all the DNS entries for the domains to point to the new IP(s)
    7. Wait for DNS propogation again (during this time places that have the new IP address already will go direct to the new server, and those that don't will go to the old dedicated IP and get redirected silently - either way people see the copy at the new location)

    PART 3:
    8. After giving enough time for things to settle, release the dedicated IP on the old server and delete anything else that is left of the account there.



    Can anyone think of any reason why the above won't work?

    I know there is some manual leg-work involved, but it is only for a small number of accounts and domains.

    And I'm aware that if the DNAT redirection works the way I'm thinking, people in locations that don't have the updated DNS records yet would experience higher latency (due to request packets going to the old server & being redirected to the new, and response packets taking the return trip via the old server too) - but this won't be a big issue for just the DNS propagation time (which I can minimise with TTL tweaks anyway).
     
  2. dspillett

    dspillett Active Member

    Joined:
    Oct 2, 2005
    Messages:
    26
    Likes Received:
    0
    Trophy Points:
    1
    Actually, I've just done little test and found the first problem: once given a dedicated address, sites are no longer available on the shared IP.

    This means I'm probably just better off:
    1. Setting TTLs very low
    2. Waiting for the old TTLs to expire
    3. Do the transfers
    4. Return TTLs to normal
     
  3. nickp666

    nickp666 Well-Known Member

    Joined:
    Jan 28, 2005
    Messages:
    770
    Likes Received:
    2
    Trophy Points:
    18
    Location:
    /dev/null
    You will have problems with the TTLs anyway, I wouldnt rely on leaving those short as a solution - ISP's such as BT cache DNS records for 72 hours regardless of the TTL on the records (I know, useless penny-pinching bastards trying to save bandwidth - you wouldnt think it would matter to a company of their size)
     
  4. dspillett

    dspillett Active Member

    Joined:
    Oct 2, 2005
    Messages:
    26
    Likes Received:
    0
    Trophy Points:
    1
    I knew that some ISPs impose an artificial minimum TTL or ~24 hours - but as much as 72? I'd call that "stupid" rather than "penny-pinching".

    As that is out of my control, and the external IPs breaking the rules, I'll just go with the "tell them to fix their systems" line. I know users could walk away in response to that, but I'm winding up my paid hosting operation anyway (these are the last few users, who had paid up for six months)...

    Is there an easy way (by easy I mean "without hand-editing apache and exim config files") of making cPanel listen for sites on the shared IP as well as the dedicated ones? Or making it listed for *every* domain on *every* IP? If that were possible then my first idea could be made to work (even accounting for the likes of BT ignoring RFCs). There aren't any HTTPS setups or other such to worry about.
     
  5. eNetHosts

    eNetHosts Well-Known Member

    Joined:
    Apr 21, 2007
    Messages:
    195
    Likes Received:
    0
    Trophy Points:
    16
    Our ISP is BT and we have 8mbps, never has supposed "caching" interfered in what we do. Especially with DNS, our request streams through perfectly. BT may cache pages; but the cache version is checked compared to the existing - if necessary and when caching could be enabled like peek times to provide a faster service. If BT cached everything everyone would migrate to another ISP due to unreliability.

    BT has a dynamic system at all their exchanges which is controlled by a backend RAS system. If you have a cached problem then you should contact BT's High Level London Team, New Customer Care Team or UK Technical Support Desk to get their systems checked out because they ain't working properly. Just dial the usual and ask to speak to any of them.

    Always clear your Local Cache's and Restart before you consider the possibility of ISPs caching. If you still have a DNS issue, run a DNS checker for resolvement and If you are on BT and do have the issue with DNS not displaying to you locally - we can take a look at the DNS and whether we get the up site. Not a problem, just ask.

    (BT usually have some problem, so maybe you do have a problem but caching by BT for 72 is unlikely, 6 hours would be more likely if even enabled, which like I said the RAS and Dynamic System should be checking for updated versions.)
     
Loading...

Share This Page