[Seminole]

Originally Posted by solefald

Actually... probability of a DOS attack is less likely than a RAID5 drive failure on the db server. Performance tanks just like we experienced here.

However, if this is true,which we probably will never find out, i want to ask the admin why the hell is the DB server running on top of RAID5?

[punky8d]

Originally Posted by solefald

Actually... probability of a DOS attack is less likely than a RAID5 drive failure on the db server. Performance tanks just like we experienced here.

However, if this is true,which we probably will never find out, i want to ask the admin why the hell is the DB server running on top of RAID5?

How did you guys know that their server is even running on Raid 5?
I don't think a failure drive on a raid 5 configuration will affect anything since it has redundancy..
We're all in a guessing game now until an admin posts something of what's going on...

i think someone's using the server to watch porn sites because of too much exposure on this thread:
http://www.e90post.com/forums/showthread.php?t=320909

[Sara]

Same thing is happening to me! "You must wait 15 seconds more to post again" and it hasnt even posted my first post. WTF

[remmib]

Yes, something is not right with this site today.

[cyberkaa]

It was double posting all day. Seems to be ok now.

[solefald]

Originally Posted by punky8d

How did you guys know that their server is even running on Raid 5?
I don't think a failure drive on a raid 5 configuration will affect anything since it has redundancy..
We're all in a guessing game now until an admin posts something of what's going on...

Oh, that was nothing more than a wild guess. However, all symptoms point towards a database issue, at least this is what i would suspect. Redundancy of the database servers does not mean shit, if the head node dies, unless this site runs on some kind of enterprise-grade database server.


http://uptime.netcraft.com/up/graph/...ww.e90post.com

But judging by ^ this url, e90post runs Apache 2 on top of Linux and I am pretty certain that db back end of this site is mysql, which is the most popular linux db server software.

mysql has pretty much 0 fault tolerance. even with clustered mysql5 setup, if your head node dies, your entire cluster falls on its face. mysql4 replication is a complete joke and it is more of a backup "oh shit" solution, than an actual redundancy.

Why do i think it is a RAID5 or RAID in general?
Because this site has 77,000 users so the database server has to be pretty beefy to handle the massive load. Running something like this on a single drive would be equal to a suicide, so I am willing to bet obscene amounts of money that RAID is used. Why RAID5? Again, because it s a popular solution, if you will. Why do i think its RAID5 vs RAID10 (RAID0 and RAID1 are out of the question completely, as well as other less popular choices.)? Because a dead drive in your RAID5 bring your entire RAID array to a crawl.

"Database Error" messages i received around 5pm further prove my point.


All of the above is my guess, however

[choicez]

i feel like i'm back on a dial-up connection.

[punky8d]

Originally Posted by solefald

Oh, that was nothing more than a wild guess. However, all symptoms point towards a database issue, at least this is what i would suspect. Redundancy of the database servers does not mean shit, if the head node dies, unless this site runs on some kind of enterprise-grade database server.


http://uptime.netcraft.com/up/graph/...ww.e90post.com

But judging by ^ this url, e90post runs Apache 2 on top of Linux and I am pretty certain that db back end of this site is mysql, which is the most popular linux db server software.

mysql has pretty much 0 fault tolerance. even with clustered mysql5 setup, if your head node dies, your entire cluster falls on its face. mysql4 replication is a complete joke and it is more of a backup "oh shit" solution, than an actual redundancy.

Why do i think it is a RAID5 or RAID in general?
Because this site has 77,000 users so the database server has to be pretty beefy to handle the massive load. Running something like this on a single drive would be equal to a suicide, so I am willing to bet obscene amounts of money that RAID is used. Why RAID5? Again, because it s a popular solution, if you will. Why do i think its RAID5 vs RAID10 (RAID0 and RAID1 are out of the question completely, as well as other less popular choices.)? Because a dead drive in your RAID5 bring your entire RAID array to a crawl.

"Database Error" messages i received around 5pm further prove my point.


All of the above is my guess, however

It is mysql.. vBulletin runs on mysql only.. And yes they definitely have RAID..No doubt about that. I'm not talking about redundancy of the database.. I don't even think they even have one.. I'm refering to the redundancy of their drives since I assumed that their in Raid..

Oh well.. At least everything's fine now..

[solefald]

Originally Posted by punky8d

I don't think a failure drive on a raid 5 configuration will affect anything since it has redundancy

Originally Posted by punky8d

I'm refering to the redundancy of their drives since I assumed that their in Raid..

I guess you don't have much experience with RAID...
Yes, RAID5 is redundant by definition, but only in terms of data safety, not performance. 1 failed drive in your RAID5 setup brings the entire RAID into a "DEGRADED" state. Degraded RAID5 comes at a price of a significant performance drop and will stay that way until you swap the failed drive and allow some time for it to rebuild.

http://www.cyberciti.biz/tips/raid5-...rformance.html

RAID 5 costs more for write-intensive applications than RAID 1.
RAID 5 is less outage resilient than RAID 1.
RAID 5 suffers massive performance degradation during partial outage.
RAID 5 is less architecturally flexible than RAID 1.
Correcting RAID 5 performance problems can be very expensive.

RAID10 (Striped mirror) is a better choice for a db server, due to faster read/write times and unless you have multiple disk failure, your performance stays pretty much intact.

Why do people chose RAID5 over RAID10? Because of the initial cost savings.
With RAID10 you lose more capacity.

RAID10: 6 x 1Tb disks = 3Tb usable space (or 2Tb if you chose double mirroring)
RAID5: 6 x 1Tb disks = 5Tb usable space [without hot spare] / 4Tb [with hot spare]

[punky8d]

Originally Posted by solefald

I guess you don't have much experience with RAID...
Yes, RAID5 is redundant by definition, but only in terms of data safety, not performance. 1 failed drive in your RAID5 setup brings the entire RAID into a "DEGRADED" state. Degraded RAID5 comes at a price of a significant performance drop and will stay that way until you swap the failed drive and allow some time for it to rebuild.

http://www.cyberciti.biz/tips/raid5-...rformance.html

RAID10 (Striped mirror) is a better choice for a db server, due to faster read/write times and unless you have multiple disk failure, your performance stays pretty much intact.

Why do people chose RAID5 over RAID10? Because of the initial cost savings.
With RAID10 you lose more capacity.

RAID10: 6 x 1Tb disks = 3Tb usable space (or 2Tb if you chose double mirroring)
RAID5: 6 x 1Tb disks = 5Tb usable space [without hot spare] / 4Tb [with hot spare]

I'm aware that you need to re-build the new hard drive that you swapped.. I understand that the performance is affected.. The amount of noticeable performance pretty much depends on how much hard drive do they have and how much activity their server performs.. I believe you just worked on small systems or with a limited amount of HW..

Until now it sometimes takes long to post and they pretty much did maintenance on it the other day..
I guess it's still RAID issue huh?? lol

Great ing capability.. I guess you're the pro and too limited to just RAID issue..
No longer want to compare credentials here..
brotha..

[solefald]

Originally Posted by punky8d

Until now it sometimes takes long to post and they pretty much did maintenance on it the other day..
I guess it's still RAID issue huh?? lol

I am a pro, but analyzing someones outage without having any back-end access is pretty much impossible. What I posted here, was about what I thought happened on the *first day*. However, I outright stated that this was an issue with database access (whatever the underlying problem is), and what I said still stands valid.

[NoKids]

i waited 15 minutes to post that i'm pissed off.

[NoKids]

i waited 15 minutes to post that i'm pissed off.

[NoKids]

i waited 15 minutes to post that i'm pissed off.

this will probably get triple posted too.

[punky8d]

I have to say lol for your 3 consecutive posts...
Everything's fine now i guess..