3:25 Am

Overheal

Ive noticed after a couple of these nights ive been having lately that boards seems to go non-responsive at some point between 3 and 4 am. Tonight this was noticed on the front page were the 2 latest posts were at 3:25 and when I managed to reonnect at 3:40 after receiving a Database Error.

Does the server just undergo a scheduled reboot or something?

weeder

its a backup

dr.bollocko

+1. I have noticed this every evening recently. a Lack of responsivity and then a datastore error. I was just server timeouted for three threads I was trying to open and was unable to re-connect to boards for about 10 minutes.
It sucks for all us boards users who keep antisocial hours or live in a different timezone and need some trolltime to pass through friday afternoon at work.

dr.bollocko

Surely a differential backup wouldnt take 20 minutes if its run every night?

orestes

It's the back up. It's the info from the day being put away so that if the site should happen to go arse ways for some reason then only one day of info is lost. It only takes five or ten minutes these days don't it?

Xavi6

If I recall correctly it used to go down for the whole hour between 3 and 4am so 15 mins is a God send. It coincides with when I take my lunch!

randylonghorn

Xavi6 wrote: »

If I recall correctly it used to go down for the whole hour between 3 and 4am so 15 mins is a God send. It coincides with when I take my lunch!

Well, I don't remember anything like an hour, but before the last major upgrade to the server, the back-up used to take ~25 mins. it's down to about 10-13 mins now, depending on the day.

seamus

How have you been here for 18 months and not known the answer to this question?

Sherifu

randylonghorn wrote: »

Well, I don't remember anything like an hour, but before the last major upgrade to the server, the back-up used to take ~25 mins. it's down to about 10-13 mins now, depending on the day.

Ya, that's what it used to be, 30 minutes tops.

tbh

seamus wrote: »

How have you been here for 18 months and not known the answer to this question?

my thoughts exactly.

Overheal

seamus wrote: »

How have you been here for 18 months and not known the answer to this question?

Sleep? I dont know. Never cared to figure it out until now. I just knew datastore errors were a regular occurence; just another character trait of the site.

dr.bollocko wrote:

Surely a differential backup wouldnt take 20 minutes if its run every night?

We do post a lot of ****e.

AlmightyCushion

dr.bollocko wrote: »

Surely a differential backup wouldnt take 20 minutes if its run every night?

I think regi said once that it's a full backup, not a differential. It means they can recover the data a lot quicker. The thing only last about 10 minutes so it's not too bad.

seamus

AlmightyCushion wrote: »

I think regi said once that it's a full backup, not a differential. It means they can recover the data a lot quicker. The thing only last about 10 minutes so it's not too bad.

Indeed. A differential backup would be quicker. But if the database goes down, then you have to restore the last full backup, then each differential backup in succession. Since a full backup is only a tiny bit of downtime, there's no problem.

Differential backups are good when you have a tonne of data and it doesn't change very often. Or just when you have a tonne of data (like, 100GB).

Overheal

I dont understand servers or networks too well, but cant a differential backup be included into a full backup? The backup server is just sitting there idle 95% of the time: surely it can cross some t's and dot some i's while it waits for its next scheduled backup?

seamus

Overheal wrote: »

I dont understand servers or networks too well, but cant a differential backup be included into a full backup? The backup server is just sitting there idle 95% of the time: surely it can cross some t's and dot some i's while it waits for its next scheduled backup?

No, it doesn't quite work like that. Regi did mention the idea of using the old server as a mirror server at some point. This could reduce or eliminate the downtime of the production server. But again, since it's 10-15 minutes at 3am, it's not exactly high on the priority list.

Overheal

Is there any way you can change how the server responds during the 15 minutes? For example instead of quickpost continuously trying to send the postdata couldnt you display an error such as "We're backing up our database. Please try again in [t]minutes [t]seconds." Just like it warns you when you are posting more frequently than the board allows, please try again in 9 seconds, etc.