corner corner (short) downtimes and / or temporary performance decrease ahead

Forum | << BookCrossing Site Announcements | << Parent Thread | Refresh | Search

corner corner

Sort Options Toggle Stats Options


Profile Image
Thanks for the heads-up, and all the work! Should we tally any outages we experience, or would that not be useful for the purpose of these trials?

Complete Thread

Profile Image
During the next days we're going to test different new server setups. Eventually this shall help us (significantly) decrease monthly costs *and* increase overall performance.

While we switch servers or server configurations there *may* be short downtimes. We expect them to last maybe minutes or just seconds (or not even that in some cases). If we should expect longer downtimes we'll try to announce them in this thread and/or set up an "Oh Bother!" information page.
No serious data loss is expected but in very rare cases a release or registration you're just working on may fail in the middle (well, nothing new here, right? ;-/ ).

Some of the configurations we're testing may show poor performance (after all, that's the most important thing we're going to check out). Slow server response and possibly an increased number of "Time Out" errors. Of course, if some test case fails severely we'll revert as instantly as possible.
We're doing performance tests before we go live, however the most relevant information is how the system responds under real live load. It's impossible to test this "in vitro".

== Background

Our current servers were brought into service in 2008, that's almost 6 years ago. They were rather high-end back then and their pricing still is quite significant. We have a monthly server bill pretty well (or "pretty badly"?) in the 4-digit numbers. Traffic is paid separately. And since we very recently experienced a very significant increase in traffic (which caused very significant additional costs) mostly due to highly increase search engine spider activity (something we cannot do much about) we have to expedite this move, that already was planned to happen later this year.

Oh, by the way: Did you notice that since a few weeks there's a separate "Donate" button on the very bottom of each page? ;-) And of course you can always support BookCrossing through our Supply Store: http://secure.bookcrossing.com/

Profile Image
And I'm about to use the Donate button : )
Profile Image
the PayPal page comes up asking how many Euros I want to donate. I can't figure out how to change it to dollars...
Profile Image
the PayPal page comes up asking how many Euros I want to donate. I can't figure out how to change it to dollars...
Ooops, I'll try to adjust this.
This happens because these donations go into our EU PayPal account (instead of the U.S. account) to avoid additional costs for moving funds to EU (where many bills - not currently the server costs - have to be paid, but only little income comes in).

For the time being:
- Very roughly one US-Dollar is about 0.73 €uros ($10 = € 7.33 || € 10 = $13.65).
- Feel free to use the "classical" way to donate through the store, if you feel more comfortable then!

In any case: A huge Thank You!
Profile Image
- Feel free to use the "classical" way to donate through the store, if you feel more comfortable then!


Done
Profile Image

Thanks!

...for the info and all the work behind the scenes. :)

Edit:
Glad to see some numbers mentioned, as most people probably do not realize what a money guzzler a big site like this can be. :-/

The bigger the community, the better, but costly for sure.
Off to use the link! :D
Profile Image
Thanks for the heads-up, and all the work! Should we tally any outages we experience, or would that not be useful for the purpose of these trials?
Profile Image
Should we tally any outages we experience, or would that not be useful for the purpose of these trials?
Since outages should only happen while we switch configurations that won't help much, methinks.

But if you feel like you're experiencing significant changes in performance (for the worse or even for the better) that would be interesting.

Thanks for mentioning this!
Profile Image
But if you feel like you're experiencing significant changes in performance (for the worse or even for the better) that would be interesting.


I don't know if you started this yesterday, but yesterday the site would not let me register a book with a pre-numbered label, despite many, many tries. I gave up and came back today, when it worked just fine and let me register several.
Profile Image
Thanks for the heads up!
Profile Image
Surprisingly I can't see (very) significant slow-downs, currently.

I say "surprisingly" because currently the web server is located in Europe while the database server still is in the USA. And these two have to communicate *permanently* and quite a lot.

However, this first test will probably only last a few hours. Well, let's see.
Profile Image
... and reverting to previous set up over the next minutes or hours (redirecting of servers is done via updating DNS servers and those may lag for a few hours or up to a day, depending on your internet provider).

For an unknown reason releases showed an error message - but were properly recorded in the database nonetheless.
Apart from that I found no uncommon and persistent performance issues during this first, fast test, so far.
Profile Image
I noticed some slowdowns accessing the forums - longer-than-usual timeouts, that kind of thing. [Also noticed that I had to log in again, both when the test began and after it finished. FWIW.]
Profile Image
The forums were painfully slow. And some forums wouldn't load at all. And I have a fast connection & a fast computer.

Profile Image
I've been seeing reports over on Facebook, today, after the test, that people can't log in. Is it related?
Profile Image
I've been seeing reports over on Facebook, today, after the test, that people can't log in. Is it related?
If you visited the BookCrossing site during the test you needed to login even if you had been logged in previously. And after the test once again.
If you didn't visit the site during the test at all you should not be affected at all.

That's because the old and the new server don't recognize each other's login cookies (and as soon as you login into one of them the login cookie for the other one is erased). But if you never touched the new server your login cookie for the old one stays intact.
Profile Image
f you visited the BookCrossing site during the test you needed to login even if you had been logged in previously. And after the test once again.
If you didn't visit the site during the test at all you should not be affected at all.

That's because the old and the new server don't recognize each other's login cookies (and as soon as you login into one of them the login cookie for the other one is erased). But if you never touched the new server your login cookie for the old one stays intact.


Understood, but I'm still being logged out on Safari and Firefox, when I shut off the computer, even if I'm still logged in when I'm powering down. Chrome stays logged in.
Profile Image
but I've been trying to reply to a PM off and on for 24 hours or more - and get an error message every time I click on the use this link to reply link!
Profile Image
but I've been trying to reply to a PM off and on for 24 hours or more - and get an error message every time I click on the use this link to reply link!


I tried to send a PM and got an error message, but got my copy. Also, got an error trying to register a book, but when I went to my shelf it was there.
Profile Image
but I've been trying to reply to a PM off and on for 24 hours or more - and get an error message every time I click on the use this link to reply link!
Does the username you want to reply to have some special character in it? In such case it might happen that the reply link is cut off just before that special character which might result in a non-existant user name.
Profile Image
but I've been trying to reply to a PM off and on for 24 hours or more - and get an error message every time I click on the use this link to reply link!
Does the username you want to reply to have some special character in it? In such case it might happen that the reply link is cut off just before that special character which might result in a non-existant user name.

My apologies! Further investigation shows she's changed her screen name since writing the PM I'm trying to reply to! No surprise the site didn't recognise her under her old name! Sorry!!
Profile Image
Last night I got a PM from one BCer with the subject line and send-message link featuring the screen name of a different BCer entirely. The bookshelf link under the sender's signature did include the correct bookshelf link, which only made the whole thing weirder. The timestamp was 9:19 PM on May 30th, Eastern Daylight Time (UTC -4). [I can supply the PM text - or forward the entire message, if that would be helpful.]

In this case there was no harm done, but if the random (?) BCer's name that had appeared on the subject line had been from, say, a known spammer, I might have deleted the note unread. And since the send-reply link was also wrong, if I hadn't noticed that and clicked on it anyway I'd have replied to the wrong person. Whether this is due to the testing or to some other site glitch, it'd be nice if it didn't happen any more {wry grin}.
Profile Image
Sorry, an error occurred
Error Details

Error Message: We're sorry. An error occurred while processing your request. Please try again, and if the error continues contact our support staff. Thank you for you patience while we clean out the bugs.

Are you sure you want to delete this item? It cannot be undone.