Hi,
We are running ESX 3.5, VC 2.5 and VCB 1.1
We have 57 VMs that undergo a full VM backup nightly using VCB
1) We have a VM (windows 2003, MS SQL 2005) running SQL, I have read other posts saying it is important to disable the SYNC driver. I have a couple of questions arround this.
1a) Why is this so - can it / does it cause corruption in the live database, also would a restored database be consistent
1b) Is this a reccomendation by VMware or is it based on ancedotal evidence
2) As an aside maybe people reading can help me with this, we ocasionally get this error (on 3 or 4 random VMs)
"snapshot creation failed: Custom pre-freeze script failed" - we don't use any custom scripts
1a) Why is this so - can it / does it cause corruption in the live database, also would a restored database be consistent
1b) Is this a reccomendation by VMware or is it based on ancedotal evidence
The problem is caused by the quiesce suspending the writes to disk as it attempts to flush the buffers in memory. If it cant do this fast enough, applications with high I/O such as Exchange can get upset. Of course even if the process does complete there is no guarentee the the database is consistant from a transactional viewpoint as VMTools has no visibility of this. The recommedation is to run pre-freeze scripts to move the application into a consistant state prior to the snap.
2) As an aside maybe people reading can help me with this, we ocasionally get this error (on 3 or 4 random VMs
"snapshot creation failed: Custom pre-freeze script failed" - we don't use any custom scripts
This sounds like a bug with VMTools. Have you opened a case with VMware ?
http://communities.vmware.com/message/890786;jsessionid=2958AFBEEE32BB675E6E4AE8DD558EDF
Did you ever find any answers on this? I have the same concern as we are just rolling out VCB and have a few SQL 2000 servers to back up. My plan is to do weekly image level backups for bare metal restores and nightly SQL dumps which we will then backup with file level backups. Am I going to run into some problems if I don't disable/remove the SYNC driver...? From what I've seen the problem may be around the DB going offline when the file quiesce takes place from that SYNC driver but haven't figured it out for sure yet.
I think there are some cases that ms sql and oracle servers will stop or fail due to snapshot. It happened ususally when the database is under heavy loading. I don't disable the sync driver of my sql servers. And vcb works fine so far. Knock, knock. But my sql servers are not heavy loaded. You have to choose between os-consistent backup with th risk of failure of database process or crash-consistent backup. Hope this helps.
1a) Why is this so - can it / does it cause corruption in the live database, also would a restored database be consistent
1b) Is this a reccomendation by VMware or is it based on ancedotal evidence
The problem is caused by the quiesce suspending the writes to disk as it attempts to flush the buffers in memory. If it cant do this fast enough, applications with high I/O such as Exchange can get upset. Of course even if the process does complete there is no guarentee the the database is consistant from a transactional viewpoint as VMTools has no visibility of this. The recommedation is to run pre-freeze scripts to move the application into a consistant state prior to the snap.
2) As an aside maybe people reading can help me with this, we ocasionally get this error (on 3 or 4 random VMs
"snapshot creation failed: Custom pre-freeze script failed" - we don't use any custom scripts
This sounds like a bug with VMTools. Have you opened a case with VMware ?
http://communities.vmware.com/message/890786;jsessionid=2958AFBEEE32BB675E6E4AE8DD558EDF
So is it reasonable to run the pre and post freeze scripts (probably just a few net stop/net start statements) and leave the SYNC driver in place to also quiesce the file system?
Thats exactly it. There are some examples of these scripts on the forum if you shop around.
Hola a todos:
Yo tenia un servidor con Windows 2000 sp4 con Oracle 9.0 en Wmware 3.5 y en el visor de eventos cada vez que aparecia LGTO_Sync la maquina se reinicia con un crush/dump.
He seguido vuestras indicaciones de desabilitar el dispositivo Sync_driver posteriormente he reiniciado y fin del problema.
Cmo dicen por aqui " Mano de Santo"
Muchas gracias a todos
SAN