| Author |
Topic |
|
ravilobo
Master Smack Fu Yak Hacker
1184 Posts |
 Posted - 2004-05-24 : 12:21:22
|
| Someone has deleted my production DB at 5pm. (lets say..).I need to recover everything till 4.59pm...what are my options...I think RAID will not help because it is for HD failure.....------------------------I think, therefore I am |
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2004-05-24 : 12:27:29
|
| You would restore the transaction log to 4:59.Tara |
 |
|
|
surefooted
Posting Yak Master
188 Posts |
Posted - 2004-05-24 : 13:29:11
|
| If it wasn't a total HD failure, you will want to backup the transaction log first if you can. If you couldn't, you will only be able to retore up to your most recent full/diff/tranlog(whichever is more recent) backup.-JonJust a starting member. |
|
|
|
ravilobo
Master Smack Fu Yak Hacker
1184 Posts |
Posted - 2004-05-25 : 02:21:29
|
| I know these options. What i am trying to say is HIGH AVAILABILITY. Suppose i have yesterdays full backup. And todays 2 pm log backup. But my db has been crashed at 5.00 pm.I know i can restore my db as of 2pm. But i will not be able to restore the transactions made in 2-5pm. Suppose mine is a BANKING databse, where even a single transaction is very important. How to make sure that i will not lose even a single transaction in the above scenario..------------------------I think, therefore I am |
|
|
|
AndrewMurphy
Master Smack Fu Yak Hacker
2916 Posts |
Posted - 2004-05-25 : 06:13:00
|
| Replication and/or log-shipping to a hot-site sounds more like what you need. |
|
|
|
mohdowais
Sheikh of Yak Knowledge
1456 Posts |
Posted - 2004-05-25 : 07:47:58
|
...and that will not come cheap. Both replication and log-shipping require at least two servers with Windows 2000 Advanced Server and SQL Server Enterprise edition. If you have a truly mission-critical database, you definitely need to think of clustering, replication (or log shipping), frequent transaction log backups, etc. But if availability is not as critical, you might find it difficult to get management to shell out the sort of money it requires. Unless you are very lucky  OS |
|
|
|
ravilobo
Master Smack Fu Yak Hacker
1184 Posts |
Posted - 2004-05-25 : 12:12:57
|
| Replication/Logshipping will not solve my problem. The requirement is not to loose a single transaction. In case of replication i will not get the non synhronised records which i lost at the time of crash. Same applies to log shipping also..Clustering..?!!Will it solve DB crash. I feel clustering is for application failure...(Money is not a criteria. My company is ready to spend money..)------------------------I think, therefore I am |
|
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2004-05-25 : 12:19:42
|
| If you've got the money, then...We run Windows Data Center with SQL Server Enterprise Edition. We've got a Compaq EVA SAN for the drives where the databases are stored. Redundant everything in the infrastructure. Clustering, replication, and log shipping. Replication though not for high availability. We worked closely with HP to setup our environment. I would suggest you select a vendor and do the same. We run HP DL760s. There is always going to be the possibility of a lost transaction. Always. With the redundancy, high availability, and everything that we've got, we minimize the amount of lost transactions.Tara |
|
|
|
derrickleggett
Pointy Haired Yak DBA
4184 Posts |
Posted - 2004-05-25 : 20:17:56
|
| Clustering..?!!Will it solve DB crash. I feel clustering is for application failure...Then you are wrong. Clustering is for server failure. That's it. Nothing more.Hey Tara, we're running Enterprise Edition (I'm Jealous) on EMC CX400s with HP DL760s. We also have six MSA1000s. How do you like the EVA series? We almost purchased one a few months ago.MeanOldDBAderrickleggett@hotmail.comWhen life gives you a lemon, fire the DBA. |
|
|
|
MichaelP
Jedi Yak
2489 Posts |
Posted - 2004-05-25 : 21:00:21
|
| You've got more than one CX400? Wow!How much data do you have on it?We've got a CX200 with not too much data on it, but we love it!The MSA1000's don't use the cache on the SCSI cards since they are in the server and not in the disk array right?Michael<Yoda>Use the Search page you must. Find the answer you will.</Yoda> |
|
|
|
derrickleggett
Pointy Haired Yak DBA
4184 Posts |
Posted - 2004-05-25 : 21:15:13
|
| We have only one CX400. I'm smoking crack and dreaming of what might be. lolOn the CX400, I have 2 enclosures full of the 146gb drives. We're still setting up the reporting and staging environments on it. We'll have about 900gb of it used I guess. Right now, we're only using around 400gb.The MSA1000s are an entry level fiber SAN. They use the Ultra-SCSI 320 drives though. There are no SCSI cards, so I think you could possibly be referring to the MSA500, which is a SCSI SAN. ??MeanOldDBAderrickleggett@hotmail.comWhen life gives you a lemon, fire the DBA. |
|
|
|
ravilobo
Master Smack Fu Yak Hacker
1184 Posts |
Posted - 2004-05-26 : 02:47:11
|
| Thank you guys for the information..------------------------I think, therefore I am |
|
|
|
MuadDBA
628 Posts |
Posted - 2004-05-26 : 08:50:13
|
quote:
Originally posted by derrickleggettClustering is for disk array failure. That's it. Nothing more.
Am I smoking silly weed? I thought clustering was to protect against CPU failures, as both nodes of a cluster use the SAME disk array, so clusters don't offer any protection against a disk array failure...your disk array has to take care of that in some other way.By the way, we just installed a compaq EVA SAN here in January. It is FAST and we are very pleased with it. Allowed us to consilidate 10 standard edition DB servers running HP RackStorage disk arrays with 1 Enterprise edition server which isn't even breaking a sweat. |
|
|
|
derrickleggett
Pointy Haired Yak DBA
4184 Posts |
Posted - 2004-05-26 : 09:01:16
|
| No. I'm smoking crack again. I've been doing this a lot lately. Probably from posting when I shouldn't. Try this.Clustering is for server failure. That's it. Nothing more.Clustering still does not cover application failure or DB issues. It just allows you to failover from one server to another if your primary server goes down. And, now I'm going to go shoot myself. Tara, do you have one of those sniped graphics you can post on here?MeanOldDBAderrickleggett@hotmail.comWhen life gives you a lemon, fire the DBA. |
|
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2004-05-26 : 12:28:40
|
Do you mean this one:  Just type this in without the spaces:[ s n i p e d ]Tara |
|
|
|
gercr
Yak Posting Veteran
53 Posts |
Posted - 2004-05-26 : 14:33:00
|
| here we use a HP EVA 5000 and after a BIG problem we have a time ago, it work perfect!!! |
|
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2004-05-26 : 14:36:31
|
| Derrick, just noticed you asked about our experience with EVA. So far it's worked out very well for us. It is super fast although the speed comes at a big price. The only problem that we have had has been due to server admin fat fingers when configuring the drives. At one point, we had the data drives from one server appear on another server. This caused a major problem obviously. After we all pulled out our hair restoring and fixing things, we recovered fine.Tara |
|
|
|
derrickleggett
Pointy Haired Yak DBA
4184 Posts |
Posted - 2004-05-26 : 18:31:44
|
| Yeah, that causes some major corruption issues. Especially if someone decides to "take a look" at the new drives on the other server. We found that out the fun way also. Did you realign the disk beginning allocation size when you created the LUNs the first time? I'm just curious how many people are doing this.MeanOldDBAderrickleggett@hotmail.comWhen life gives you a lemon, fire the DBA. |
|
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2004-05-26 : 18:42:14
|
quote:
Originally posted by derrickleggett
Did you realign the disk beginning allocation size when you created the LUNs the first time? I'm just curious how many people are doing this.
I don't have any idea.  The server admins take care of that stuff for us.Tara |
|
|
|
gercr
Yak Posting Veteran
53 Posts |
Posted - 2004-05-27 : 17:37:28
|
| a EVA TIP. If you have a group of 8 disk (the minimun permit) always left a free space equal to a 1 disk of the group. This will avoid you a lot of problems. |
|
|
|
rsj2004
Starting Member
1 Post |
Posted - 2004-06-23 : 08:37:00
|
| Hi ravilobo,I am also facing the same problem. As the replies here mention that clustering is just for server failure. That is if one fails, other takes over. Replication does the immediate copying of data on the replicated databases.But the ultimate the question remains. How to get those transaction between 2pm to 5pm. ? in that time due to failure of one DB the records were not updated to that DB by replication.Have found any answer ? I am in the same problem.Please help |
|
|
|
Next Page
|
High availabilty