marrowyung
asked on
data conflict of Oracle RAC
hi,
When Oracle RAC detect conflict of data, how it handle it ?
any email /alert will send out ?
data gone ?
When Oracle RAC detect conflict of data, how it handle it ?
any email /alert will send out ?
data gone ?
>> When Oracle RAC detect conflict of data, how it handle it ?
I am not sure what you mean by conflict of data. RAC is a mechanism to allocate/utilize resources for scalability. However the database is a database. Oracle's read consistency and transaction isolation is not compromised due to RAC.
I am not sure what you mean by conflict of data. RAC is a mechanism to allocate/utilize resources for scalability. However the database is a database. Oracle's read consistency and transaction isolation is not compromised due to RAC.
1. If two RAC installation are installed to place Data in common Disk Groups (by error) of the ASM only one RAC installation will be able to run. The Disk groups should be different.
2. If you export data with the Data Pump the dump files repository should be visible to all nodes. If this is not the case use the parameter:
CLUSTER=N
2. If you export data with the Data Pump the dump files repository should be visible to all nodes. If this is not the case use the parameter:
CLUSTER=N
ASKER
slightwv,
no. I studing MariaDB, it has this problem but as multi master (including bi directional replication) can write to the SAME data cell but 2 x nodes, then how oracle solve it, this is what I want to know.
Sujith,
"However the database is a database. Oracle's read consistency and transaction isolation is not compromised due to RAC."
I think this gives an hints ! only one RAC node can write to the share storage volume on the SAME cell at one time, this is what RAC will do.
schwertner,
"If this is not the case use the parameter:
CLUSTER=N"
what is that for ?
no. I studing MariaDB, it has this problem but as multi master (including bi directional replication) can write to the SAME data cell but 2 x nodes, then how oracle solve it, this is what I want to know.
Sujith,
"However the database is a database. Oracle's read consistency and transaction isolation is not compromised due to RAC."
I think this gives an hints ! only one RAC node can write to the share storage volume on the SAME cell at one time, this is what RAC will do.
schwertner,
"If this is not the case use the parameter:
CLUSTER=N"
what is that for ?
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
schwertner,
None of that really has anything to do with multi-master replication which is what the question is about.
None of that really has anything to do with multi-master replication which is what the question is about.
slight,
All of conflicts I have mentioned can arise in RAC environment. I hope the topic is not about Oracle bugs in RAC.
All of conflicts I have mentioned can arise in RAC environment. I hope the topic is not about Oracle bugs in RAC.
>>can arise in RAC
Only the CLUSTER parameter of export. Deadlocks, corrupt blocks and Dataguard are issues in non-RAC as well. They also really aren't related to data conflicts as described in the question.
All that said, the question isn't about RAC. Sure it was in the title but in #a42595205 it changed into really about replication not RAC.
Only the CLUSTER parameter of export. Deadlocks, corrupt blocks and Dataguard are issues in non-RAC as well. They also really aren't related to data conflicts as described in the question.
All that said, the question isn't about RAC. Sure it was in the title but in #a42595205 it changed into really about replication not RAC.
ASKER
slightwv,
" All replication systems need some form of conflict resolution."
MySQL multi master is based on replication OR NDB cluster, but what should I say is, MariaDB's Galera Cluster has this kind of conflict, if transaction conflict, then it rollback that transaction.
schwertner,
"The worst cases are corrupted blocks, files - but this can be a topic for big article."
yeah, another big topic ! I knew this is a problem of Oracle and that's why in Oracle 18c, it use AI to detect it and repair it! seems LInux files system problem ? MySQL on linux also has this kind of problem too.
there are also other big big problem of Oracle too in terms of security.
"CLUSTER=N is a very important Data Pump parameter for professionals working in RAC environment. "
this seems more on OS files locking ? but about data conflict, right ?
"Another Data Conflict is in the Data Guard environment - the standby DB has not received archive logs from the Primary. There are special technologies to detect the gaps and to resolve them."
it is again, related to other latency ? network latency.
" This Is the reason for Oracle to try to move toautonomous DB with Oracle 18c "
I read autonomous DB, but why it is important as I don't quite get the point.
"about replication not RAC."
wait, it is about RAC data conflict as I replied, might bit a confused somewhere.
" All replication systems need some form of conflict resolution."
MySQL multi master is based on replication OR NDB cluster, but what should I say is, MariaDB's Galera Cluster has this kind of conflict, if transaction conflict, then it rollback that transaction.
schwertner,
"The worst cases are corrupted blocks, files - but this can be a topic for big article."
yeah, another big topic ! I knew this is a problem of Oracle and that's why in Oracle 18c, it use AI to detect it and repair it! seems LInux files system problem ? MySQL on linux also has this kind of problem too.
there are also other big big problem of Oracle too in terms of security.
"CLUSTER=N is a very important Data Pump parameter for professionals working in RAC environment. "
this seems more on OS files locking ? but about data conflict, right ?
"Another Data Conflict is in the Data Guard environment - the standby DB has not received archive logs from the Primary. There are special technologies to detect the gaps and to resolve them."
it is again, related to other latency ? network latency.
" This Is the reason for Oracle to try to move toautonomous DB with Oracle 18c "
I read autonomous DB, but why it is important as I don't quite get the point.
"about replication not RAC."
wait, it is about RAC data conflict as I replied, might bit a confused somewhere.
>>wait, it is about RAC data conflict as I replied, might bit a confused somewhere.
We already answered this: RAC isn't replication and doesn't have a data conflict issue.
You cannot insert the same record in two different instances and commit them in RAC. You are bound by the same locking in RAC that you are in a single instance.
We cannot say it any clearer than we did 4 days ago.
The confusion came in when you brought up "multi master (including bi directional replication)". As I pointed out, replication isn't RAC.
We already answered this: RAC isn't replication and doesn't have a data conflict issue.
You cannot insert the same record in two different instances and commit them in RAC. You are bound by the same locking in RAC that you are in a single instance.
We cannot say it any clearer than we did 4 days ago.
The confusion came in when you brought up "multi master (including bi directional replication)". As I pointed out, replication isn't RAC.
The problem of the RAC is that the nodes lies on different boxes. But they share their RAM in the so called "RAM fussion" (or something like this!). This can't be done without super fast network. But one nasty problem is that the leading node has to know where is each record placed and when. To make things worst every computer has his own clock. Theoretically they should be synchronized, practically there are small time differences. This is very bad and we experienced difficulties to fix this. Oracle has developed a service VKTM. This is the "time keeper" - extremely aggressive process. It is on the top of all Oracle prozesses. Even on single instance installation. First thing I do is to suppress this process. So you learn that there are not only data problems, there are ... time problems also ...
"I read autonomous DB, but why it is important as I don't quite get the point."
The cloud (Amazon, Oracle, ...) gives the possibility also for small companies to apply Oracle DBs. So the number of the DB's arises in the Data Centers. But you (with your experience) understand that it is impossible to administer effectifelly this amount of DBs. There are not so much DBA in the world. Thatswhy Oracle moves to automatization and AI. But on standardized Hardware - Superclusters and Exadata produced by Oracle and in Data Centers on Oracle premises.
The cloud (Amazon, Oracle, ...) gives the possibility also for small companies to apply Oracle DBs. So the number of the DB's arises in the Data Centers. But you (with your experience) understand that it is impossible to administer effectifelly this amount of DBs. There are not so much DBA in the world. Thatswhy Oracle moves to automatization and AI. But on standardized Hardware - Superclusters and Exadata produced by Oracle and in Data Centers on Oracle premises.
ASKER
"automatization a"
ok, so move to that automatization DB means, DB manage itself ?
ok, so move to that automatization DB means, DB manage itself ?
ASKER
tks all.
If you have some information that talks about this in RAC, I'll take a look at it.