Forum Stats

  • 3,838,654 Users
  • 2,262,389 Discussions
  • 7,900,724 Comments

Discussions

Running Instances in SOA EM

Ashutosh Singh
Ashutosh Singh Member Posts: 103
edited May 14, 2013 10:13AM in SOA Suite Discusssions
Hi,
I want to understand if we have lot of running instances in SOA , what is the impact of it.
As far as I know that running instances means that
1: Running instances are in dehydrated state. They should not consume systems memory continuously to make the system slow.
2: Running instances will consume space in the database as they are nothing but a process under execution.

If above two points are correct then, Please help me in below query,
If we have too many running instances in the system say due to some non recoverable fault(other system is down) in target system and source system is sending transactional data
regularly,

Will my soa infra tablespace be filled with the same rate by which it is filled when SOA saves a completed instance ?

Thanks,
Ashutosh
Tagged:
Alena

Answers

  • Anuj Dwivedi--Oracle
    Anuj Dwivedi--Oracle Member Posts: 7,077 Employee
    Ashutosh,
    I want to understand if we have lot of running instances in SOA , what is the impact of it.
    Depends on many factors. Depends on why are they stuck in running state, how old are they and how many of them. Also depends on server's capacity as well as DB's capacity.
    1: Running instances are in dehydrated state. They should not consume systems memory continuously to make the system slow.
    It's not true always. Only if a process encounters the dehydration activity then only instances are dehydrated.
    2: Running instances will consume space in the database as they are nothing but a process under execution.
    Running instances which are having large number of retries and are failing continuously may cause severe performance issue as load increases rapidly on DB and on server both. Remember that for each fault, data will be written in DB as well as in logs (which will consume CPU and memory both).
    Will my soa infra tablespace be filled with the same rate by which it is filled when SOA saves a completed instance ?
    No, because additional data about faults will also get stored in DB along with the other details of instance (which are stored in case of normal completion). Moreover if retries are enabled and one instance is failing multiple times then issue may become severe as it will impact both SOA server and DB.

    Remember that by default "RUNNING" instances are not purged so if you are having too old and too many running instances then even regular purge may not help in reducing the size of SOA INFRA tablespace and in that case, SQL queries may also start performing poorly which will eventually impact the SOA server(s).

    So it's not a good idea to have too many running instances for a too long time specially in high load messaging systems.

    Regards,
    Anuj
  • Hi Anuj,
    Thanks for your reply.

    Will my soa infra tablespace be filled with the same rate by which it is filled when SOA saves a completed instance ?
    No, because additional data about faults will also get stored in DB along with the other details of instance (which are stored in case of normal completion). Moreover if retries are enabled and one instance is failing multiple times then issue may become severe as it will impact both SOA server and DB.

    Remember that by default "RUNNING" instances are not purged so if you are having too old and too many running instances then even regular purge may not help in reducing the size of SOA INFRA tablespace and in that case, SQL queries may also start performing poorly which will eventually impact the SOA server(s).

    So it's not a good idea to have too many running instances for a too long time specially in high load messaging systems.
    I completely agree with you with this and understand that instances goes in dehydration when dehydration type activity is called.
    However situations like this which I tried to explain where target system is not available , then also instances are visible like running in SOA EM.
    My this query is much inclined towards this situation. Also these running instances are not retrying multiple times. Hence logs are also not being written multiple times.
    They are stuck in that state.


    Then in this kind of situation, Is it possible that they fill the tablespace with higher rate ?
    or
    it will be just saved like normal instances. You can assume that they are non recoverable and even if target system gets ready in some time source has to do re transmission of those objects.

    Thanks,
    Ashu
This discussion has been closed.