Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Darth Loo-knee]

I didn't see it mentioned before, but has anyone else noticed that the server seems to go AWOL for ~5 minutes at a time every now and then?

Yeah, we know what causes it, and was aware the day before migration.

We have a couple of options available, but are still trying to get a baseline, so we know if any tweaks improve the situation or not.

Is it every time I view a page that's causing it?    

Yeah, its all you rather fault, gayboy


In layman terms:
Its the hourly backups

For those that understand, to get a point-in-time backup of the database (and it has to be point-in-time to maintain consistency), it is necessary to lock the database tables before the backup starts, and release the locks when its complete. In testing, this was taking less than 30s, which we thought we would get away with.  Now traffic has ramped up, and an unexpected high number of users still browsing the old YaBB site, the server is working much, much harder, and its taking a few minutes. People then get fed up with waiting, so constantly hit refresh, giving the server yet more requests to deal with.

That certainly is not Laymans Terms Jammie 
I for one ain't got a bloody clue what you were just bangin' on about 

[TheBoy]

I didn't see it mentioned before, but has anyone else noticed that the server seems to go AWOL for ~5 minutes at a time every now and then?

Yeah, we know what causes it, and was aware the day before migration.

We have a couple of options available, but are still trying to get a baseline, so we know if any tweaks improve the situation or not.

Is it every time I view a page that's causing it?    

Yeah, its all you rather fault, gayboy


In layman terms:
Its the hourly backups

For those that understand, to get a point-in-time backup of the database (and it has to be point-in-time to maintain consistency), it is necessary to lock the database tables before the backup starts, and release the locks when its complete. In testing, this was taking less than 30s, which we thought we would get away with.  Now traffic has ramped up, and an unexpected high number of users still browsing the old YaBB site, the server is working much, much harder, and its taking a few minutes. People then get fed up with waiting, so constantly hit refresh, giving the server yet more requests to deal with.

That certainly is not Laymans Terms Jammie 
I for one ain't got a bloody clue what you were just bangin' on about 

Sorry, the laymans bit was the backups are causing the problem.

The other bit was for those who give a monkeys

[Darth Loo-knee]

Ah... now why didn't you just say that 

[cem_devecioglu]

I didn't see it mentioned before, but has anyone else noticed that the server seems to go AWOL for ~5 minutes at a time every now and then?

Yeah, we know what causes it, and was aware the day before migration.

We have a couple of options available, but are still trying to get a baseline, so we know if any tweaks improve the situation or not.

Is it every time I view a page that's causing it?    

Yeah, its all you rather fault, gayboy


In layman terms:
Its the hourly backups

For those that understand, to get a point-in-time backup of the database (and it has to be point-in-time to maintain consistency), it is necessary to lock the database tables before the backup starts, and release the locks when its complete. In testing, this was taking less than 30s, which we thought we would get away with.  Now traffic has ramped up, and an unexpected high number of users still browsing the old YaBB site, the server is working much, much harder, and its taking a few minutes. People then get fed up with waiting, so constantly hit refresh, giving the server yet more requests to deal with.

Bringing db to single user or lock mode is a must in all databases unless a new technology is introduced.. However, although expensive there are some alternative solutions like replicated/image backup but not sure this is a must or priority now..

just my opinion, stopping the old site and giving all efforts to the new one will be easier imo..And will decrease the server activity on disks ..

And from my experience, I always prefer to back up(sql) at late midnight hours , and during the day time, write the inserts, updates and deletes on sequential log files in case of a recovery necessity..

[Darth Loo-knee]

I didn't see it mentioned before, but has anyone else noticed that the server seems to go AWOL for ~5 minutes at a time every now and then?

Yeah, we know what causes it, and was aware the day before migration.

We have a couple of options available, but are still trying to get a baseline, so we know if any tweaks improve the situation or not.

Is it every time I view a page that's causing it?    

Yeah, its all you rather fault, gayboy


In layman terms:
Its the hourly backups

For those that understand, to get a point-in-time backup of the database (and it has to be point-in-time to maintain consistency), it is necessary to lock the database tables before the backup starts, and release the locks when its complete. In testing, this was taking less than 30s, which we thought we would get away with.  Now traffic has ramped up, and an unexpected high number of users still browsing the old YaBB site, the server is working much, much harder, and its taking a few minutes. People then get fed up with waiting, so constantly hit refresh, giving the server yet more requests to deal with.

Bringing db to single user or lock mode is a must in all databases unless a new technology is introduced.. However, although expensive there are some alternative solutions like replicated/image backup but not sure this is a must or priority now..

just my opinion, stopping the old site and giving all efforts to the new one will be easier imo..And will decrease the server activity on disks ..

And from my experience, I always prefer to back up(sql) at late midnight hours , and during the day time, write the inserts, updates and deletes on sequential log files in case of a recovery necessity..

Cem, I have too say mate your English is Brilliant. I used to read some of your posts twice to sometimes understand them, but now they are perfect

You also need a Badge as big as a Dustbin Lid to understand Jammie 

[TheBoy]

Bringing db to single user or lock mode is a must in all databases unless a new technology is introduced.. However, although expensive there are some alternative solutions like replicated/image backup but not sure this is a must or priority now..

And from my experience, I always prefer to back up(sql) at late midnight hours , and during the day time, write the inserts, updates and deletes on sequential log files in case of a recovery necessity..

Plan A is replicate to another server, then back that up.

As to hourly backups, probably not that important in the scale of things. Temporarily set to a single middle-of-night backup.  Also, at last crash (8:10pm), I increase the RAM by 50%, as that was the biggest issue...

[cem_devecioglu]

I didn't see it mentioned before, but has anyone else noticed that the server seems to go AWOL for ~5 minutes at a time every now and then?

Yeah, we know what causes it, and was aware the day before migration.

We have a couple of options available, but are still trying to get a baseline, so we know if any tweaks improve the situation or not.

Is it every time I view a page that's causing it?    

Yeah, its all you rather fault, gayboy


In layman terms:
Its the hourly backups

For those that understand, to get a point-in-time backup of the database (and it has to be point-in-time to maintain consistency), it is necessary to lock the database tables before the backup starts, and release the locks when its complete. In testing, this was taking less than 30s, which we thought we would get away with.  Now traffic has ramped up, and an unexpected high number of users still browsing the old YaBB site, the server is working much, much harder, and its taking a few minutes. People then get fed up with waiting, so constantly hit refresh, giving the server yet more requests to deal with.

Bringing db to single user or lock mode is a must in all databases unless a new technology is introduced.. However, although expensive there are some alternative solutions like replicated/image backup but not sure this is a must or priority now..

just my opinion, stopping the old site and giving all efforts to the new one will be easier imo..And will decrease the server activity on disks ..

And from my experience, I always prefer to back up(sql) at late midnight hours , and during the day time, write the inserts, updates and deletes on sequential log files in case of a recovery necessity..

Cem, I have too say mate your English is Brilliant. I used to read some of your posts twice to sometimes understand them, but now they are perfect

You also need a Badge as big as a Dustbin Lid to understand Jammie 

Thanks Darth 

[cem_devecioglu]

Bringing db to single user or lock mode is a must in all databases unless a new technology is introduced.. However, although expensive there are some alternative solutions like replicated/image backup but not sure this is a must or priority now..

And from my experience, I always prefer to back up(sql) at late midnight hours , and during the day time, write the inserts, updates and deletes on sequential log files in case of a recovery necessity..

Plan A is replicate to another server, then back that up.

As to hourly backups, probably not that important in the scale of things. Temporarily set to a single middle-of-night backup.  Also, at last crash (8:10pm), I increase the RAM by 50%, as that was the biggest issue...

Plan A is    (replication alternative scenarios check may be necessary will take time)

agreed.. hourly sql backups not necessary..

few months  ago I tested (it was production actually ) ordinary SSD disks (Kingston) on a server.. although everyone was against and the marketing people recommend expensive server type SSDs (which are silly expensive) I tried 3 SSDs on a server.. 2 are mirror (boot) and 1 for data .. it was really fast and I can recommend them.. must mention read lots of papers about them.. and considerably long time is necessary to see some erroneous memory blocks.. so , you can use them for either virtual memory or other purposes

[jonnycool]

Are you still talking English Cem? 

[TheBoy]

few months  ago I tested (it was production actually ) ordinary SSD disks (Kingston) on a server.. although everyone was against and the marketing people recommend expensive server type SSDs (which are silly expensive) I tried 3 SSDs on a server.. 2 are mirror (boot) and 1 for data .. it was really fast and I can recommend them.. must mention read lots of papers about them.. and considerably long time is necessary to see some erroneous memory blocks.. so , you can use them for either virtual memory or other purposes

We run an HP SMART Array P400 disk controller, which may be a bit fussy on SSDs...  ...lot of money on something that might not work...

[cem_devecioglu]

Are you still talking English Cem? 

  I hope so

[cem_devecioglu]

few months  ago I tested (it was production actually ) ordinary SSD disks (Kingston) on a server.. although everyone was against and the marketing people recommend expensive server type SSDs (which are silly expensive) I tried 3 SSDs on a server.. 2 are mirror (boot) and 1 for data .. it was really fast and I can recommend them.. must mention read lots of papers about them.. and considerably long time is necessary to see some erroneous memory blocks.. so , you can use them for either virtual memory or other purposes

We run an HP SMART Array P400 disk controller, which may be a bit fussy on SSDs...  ...lot of money on something that might not work...

hmmm... Hp   I'm also not sure about them.. when searching they (marketing people) also confused me and tried to sell very expensive array control card and disks for SSD configuration..(I show them the door as a result)  and then I go for OEM.. no probs so far..

and one more point, I see we put real economic burden on your head

[Darth Loo-knee]

Are you still talking English Cem? 

Nope, Computer Jargon

[TheBoy]

I now have two PM,s that I need to reply to but cant.I believe Twiglet had a similar problem earlier today.Anyone know why ?
Does it apply to everyone all the time ? A glitch ?.......

its a restriction we have in place to protect the server. As time goes on, there may be an opportunity to increase this figure, but there will always be a restriction

Following today's problems, we are certainly going to have to delay an increases for a few days.

This software is still new to us, and we still have to learn all the interelationships, and how they impact performance and stability of the site. Please bear with us.

[Vamps]

What about that clock, it has to go.......nearly 8 hours on this new forum, looks bad.....