The server/site has been upgraded!

Yesterday the site was moved to a new server! Now we have 2GB of RAM (was 1GB), 8 core E5-2650L Xeon @1.8GHz 20MB cache CPU (was 4 core L5520 Xeon @ 2.27GHz 8MB cache), 48 GB or storage space (was 40GB), 4TB(!!!!) of monthly transfer (was 400GB), 10Gb/s Ethernet infrastructure (was 1Gb/s), new disks with more cache.

Not bad, huh? The most important upgrade is the RAM. One extra Gigabyte allows more possible connections and more cache (APC, pagespeed, MySQL). Not much, actually, you may not even notice the improvements (I think the latency decreased by 10-15%), but now we have more breathing space, more overload protection, more stability and almost unlimited bandwidth.

For example, as long as nobody’s ripping the whole site (I immediately block such IP addresses), the PHP content is served entirely from the APC cache (100% hit rate)!

See also Site performance, Server errors, outages and tunings

8 thoughts on “The server/site has been upgraded!”

  1. The site now hangs every time a page containing multiple images’ loaded, e.g.
    /blog/wp/archives/4437/rik-28

    You might want to check your “immediately block such IP addresses” part and/or your cache mechanism, because trying to load ONE SINGLE PAGE that has a gallery inside leads to
    “Unable to connect – Firefox can’t establish a connection to the server at http://www.likera.com.” for several hours.

    (and yes, I had to wait ~7hours before I can make this report.)

  2. * additional info –
    experienced this problem on both FFOX 20.0, WinXP and FFOX 19.0, ubuntu.

  3. wrong link, sorry: its
    /blog/wp/archives/6

    (or, basically, any page that “random images” lead me to)

  4. The network was down for 8 hours. I could not even ping the default gateway, so I reboot the system and logged a support ticket.

    Most likely your browser cached some pages so you did not see the “Host is not reachable” message.

  5. The node has just been migrated to another server. The CPU is now faster but with less cache:

    Intel(R) Xeon(R) CPU E5-2630L 0 @ 2.00GHz 15MB cache

    Let’s see if that makes any difference.

  6. interesting. you said there were two downs right?

    both of them (or, if theres more, at least two of them) “coincidentally” happened right when i try to open a gallery page, leaving a page half-loaded (that is, about 40% of all images in the page visible, other 60% blank) for me – thus I think we can reasonable deduct that these two events are linked somehow.

    I’ll see what happens if I access a few gallery pages now, and just to be clear (in case it bought the network down again): im not trying to damage your great site anyhow :), absolutely love it and i would like to see that anomaly fixed/fixes itself asap.

  7. no problem accessing gallery pages now.

    accident earlier is unlikely to be a coincidence. i dont know how the site’s set up thus i cant offer advices here, just plain fact: TWICE in last 3days, I tried to open a gallery page, php work thus the html itself loads, images begin to show up one by one, then somewhere halfway there everything stop: no further image, click on any links lead to a unable to connect error (thus i thought i got banned and its a problem w/ blocking mechanism/dos protection). My attempt and the resulting network breakdown’s very similar in both cases – Hope this helps in solving the problem.

  8. Actually, there were four outages, but two of them were planned (migration).

    The images should load when they are visible in your browser (loading on-demand). If they are hosted on this web-site, everything should work (but JavaScript must be allowed and you might need to exclude likera.com from AdBlock in Chrome, because ABP “corrupts” menu CSS).

    But some images are hosted by PhotoBucket, and this is what’s currently broken, because they changed the hotlinking mechanism. Since about two hundred posts are affected, it’s not an easy task to fix. But it’s in my todo list.

    Anyway, since you’re saying that gallery pages work now, I assume that the problem was indeed related to the outages.

Leave a Reply

Your email address will not be published. Required fields are marked *