Results 1 to 7 of 7
  1. #1
    Join Date
    Dec 2005
    Posts
    368

    What to expect with planned VPS downtime?

    I have several unmanaged Xen-based VPSes with a company that I've for the most part been very pleased with for the past couple years. I recently got an email from them saying that the physical server my VPS was on needed some components replaced and that the VPS would be "briefly unavailable" on a Saturday morning to do so.

    Now, my definition of "briefly unavailable" is a few minutes, maybe 5 minutes max. I assumed that they would migrate the VPSes to a backup physical server while they fixed the other one, then migrate them back afterwards. In fact, I thought that Xen was capable of doing this in real-time with no downtime.

    But the VPS was down for over an hour, so they obviously just shut down the box with the VPSes running. Because of the nature of my website, that particular Saturday was very inconvenient for me to have extended downtime. I was rather upset and contacted them to let them know, but they seemed rather indifferent and suggested that I look into upgrading to their redundant/failover services.

    So my question is, am I being unreasonable to expect that *planned* downtime be limited to a couple minutes, or is it normal for companies to just take the servers down while they perform the maintenance? I thought that real-time migration to a different server was one of the benefits of virtual servers and I was surprised that they didn't take advantage of it.

  2. #2
    Join Date
    Nov 2005
    Location
    London UK
    Posts
    683
    I do agree that "briefly unavailable" is about 5 minutes.

    Did they give you a maintenance window ?

    Normally we say the server will be down for 5 minutes, but allow 60 minutes in case of any problems. Taking down a server is the worst possible time for problems - sometiems they don't come back up.
    UK SSD VPS without compromise - upto 250k IOPS !
    www.clustered.net

  3. #3
    Join Date
    Oct 2005
    Location
    Six Degrees From You
    Posts
    1,075
    I would say that 1 hour for this is fine.

    Did they say that the downtime could last upto an hour?

    Paul

  4. #4
    Join Date
    Jun 2003
    Location
    UK
    Posts
    6,601
    Few things pop into mind (most from bitter experience)

    a) The host might of expected it to brief but something might of cropped up making it take longer. Plan for the worst, expect the best
    b) The time to migrate a number of VPS might take a few hours, downtime might be only 10 minutes (if things go to plan). Recent example for us is we relocated the best part of a 1000 VPS and we aimed for downtime to be about 1 hours (some people ended down for 2 but see a) ) means the benefits of migration are less than just powering the node down

    If you have a !Gbit connection and 150GB of data on the VPS node then its 25-30 minutes to transfer. If you expect a brief downtime to be 5 minutes it doesn't make sence to transfer the data

    c) Starting up VPS isn't instant. You are basically starting up 20/30 servers so it can take a good 30 minutes for everything to startup if running something like cPanel

    Rus
    Russ Foster - Industry Curmudgeon

  5. #5
    Join Date
    Sep 2006
    Location
    Cardiff - United Kingdom
    Posts
    1,568
    As a2b2 has said, maybe it taken longer than expected.

    However I do think said host should have specified how long they think the downtime *should* last - even a rough estimate (e.g. "20-60 minutes") would be better than the ambiguous "briefly unavailable" you received.

    And I agree that "briefly unavailable" does only mean <10 minutes.
    Plagiarism Guard - Protect Against Content Theft
    Tristan Perry - Personal blog

  6. #6
    Join Date
    Dec 2005
    Posts
    368
    Thanks for the feedback guys, the guidance they gave is that the maintenance would take place within a specific 4 hour window (which it did) but that individual VPS would be "briefly unavailable" so taking those two statements together I assumed that they would be down for a few minutes but it could be anywhere during that period.

    If the downtime was *expected* to take an hour, my expectation is that most hosting companies would try to do so very early in the morning (such as somewhere in the 2am-4am window) instead of on a weekend at noon which is what happened.

    Like I said I've been overall very pleased w/ the company overall but I think the one aspect that they need improvement on is communication to customers and this hasn't helped my perception in that regard!

  7. #7
    Join Date
    Feb 2004
    Posts
    371
    tobiasly,

    I would consider this a normal scenario when it comes to required maintenance on a server. The hosting company is better off, doing a one hour downtime window, that to leave it as it is and make matters worse in the long run.

    Also, Live migration is very different from regular Xen migration. Live migration (100% uptime during the move) requires some type of shared storage between servers, either FC/iSCSI. For the most part, VPS providers do not offer these services, unless they offer some type of redundant/fail over services.

    In all honest, if you have been with them the "past couple of years"
    I have several unmanaged Xen-based VPSes with a company that I've for the most part been very pleased with for the past couple years.
    Then, an hour of ANNOUNCED maintenance on your server was rather needed, and honestly, nothing wrong with it.

    Keep in mind, they could have added more RAM, or upgraded processors to the node, making your vps run even better.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •