sean
|
Hamster had a seizure?What was that about then? Critical error messages galore.
|
tahir
|
Andrew's been trying to feed the server cuppas. Told him too much caffeine was a bad idea
|
jema
|
When exactly and what messages?
|
Behemoth
|
Never, ever let that happen again. The prospect of work was most terrifying.
|
sean
|
15 minutes ago-ish. 'Cannot contact database'. Upset my email and myFF as well I think.
|
tahir
|
| jema wrote: | | When exactly and what messages? |
Critical php error about 10-15 mins ago
|
Behemoth
|
I got a big Joomla message saying no contact could be made.
|
jema
|
the mysql server when down, I restarted it this morning with more aggressive parameters. The restart on failure was automatic and hence has started again with the same problem parameters 
I think I will do another restart on it, to make things safer.
|
jema
|
restarted again. Should be safer this time.
|
Barefoot Andrew
|
I've now stopped feeding it cuppas and all manner of unimportant items.
A.
|
sean
|
And again.
|
jema
|
I know. All part of trying to find faster settings whilst staying stable.
|
Jonnyboy
|
Twas painful this morning, but I assume that was planned and known. 
|
jema
|
Certainly anticipated, currently I am trialling another backup system.
We seem to be between a rock and a hard place on performance. Tweaking certain settings leads to greater performance, but crippling backups and as we have seen a stability issue.
But plainly backups are something that must have priority. So if I don't get much further today with the new backup system, I will throttle things further back to ensure tomorrows backup is done.
|
Jonnyboy
|
Done mean to minimise your hard work. But i thought the speed was perfectly acceptable before the tweaking began.
|
Fee
|
good backups = wise 
|
sean
|
Are you a bit obsessive about backing-up for some reason Fee?
|
jema
|
By most definitions it was.
But a horde of internal database statistics suggested that it could be better, some of those issues are proving fruitful, but where our server is different to a typical web server is proving a problem.
We have far more active database tables than any typical scenario, and therefore to perform at the best we need to push table caches to out of the ordinary levels.
I tried this a few weeks back and ran into similar problems to what we are seeing now, but this time round I have been using a raft of "clever" tools to do much the same thing, but free of the risk that I was getting other settings wrong and therefore being mislead about it all being down to being unable to push the table cache higher.
Thus far though, we are hitting the same issues.
But it still may be possible to find a level where we are stable, faster and either the new backup works, or the old one will run acceptably.
Don't worry though about this being a constant process. I need the system to be reliable and hate the downtime more than anyone.
|
jema
|
| Fee wrote: | | good backups = wise |
I'm really grumbling at having to consider a changed system. The old system WORKED.
|
jema
|
Have scaled back the changes again, it does seem table locking starts to struggle badly with a large table cache, and that is what all backups tend to do.
|
jema
|
Talk about full circle. I have just really scaled back a setting to what in my naive days when I started running servers I rationlised would be right 
Several books and tons of research, then told me I was wrong and I caved into conventional wisdom.
I have now read a comment from someone hosting in a similar way that states categorically that in their practical experience I was right first time 
|
tahir
|
So are we getting better?
|
jema
|
Well we are now reset to parameters that are generally a lot more on the ball, plus the critical parameter that was screwing us, is now massively reduced to a point where I think the backups will work fast without effecting the site.
I won't apologise too much for this learning curve, as you wouldn't believe how many hours I have spent reading books and web advice telling me to do the wrong thing
|
tahir
|
| jema wrote: | | I won't apologise too much for this learning curve, as you wouldn't believe how many hours I have spent reading books and web advice telling me to do the wrong thing |
See marigold, told you "NEVER READ THE INSTRUCTION BOOK!!!"
|
marigold
|
| tahir wrote: | | jema wrote: | | I won't apologise too much for this learning curve, as you wouldn't believe how many hours I have spent reading books and web advice telling me to do the wrong thing |
See marigold, told you "NEVER READ THE INSTRUCTION BOOK!!!"
|
(wossat you've morphed into???)
|
jema
|
Well I was not online to see speed during this mornings backup, but it did complete in 3 hours which is an hours improvement on what it has been doing when things were previously working well. When things were being a problem that time could be far worse.
There is still some tweaking to do, but only around areas that will not give us any issues of stability.
|
