First of all, I appreciate the discussion that this issue has generated here.
Have had an SR open for the past week, and they still haven't confirmed whether
these services should be registered in the standby-side listener or not. <sigh>
I found the following statement in the alert log during standby startup,
confirming that all services, including those that have a role of PRIMARY are
being set.
ALTER SYSTEM SET service_names='ABCR_PERL,ABCR_TSAM,ABCR_USR,ABCR_WEB,ABCR1'
SCOPE=BOTH SID='*';
Of course I have no way to confirm whether this is expected behavior on standby
startup or not.
With our work-around (manual deregistration) on the standby database,
connection attempts work as expected. The first ADDRESS_LIST entry is
bypassed, and the second entry is used to connect to the primary database.
There are no ORA-01033 errors.
ALTER SYSTEM SET service_names='ABCR1' SCOPE=BOTH SID='*';
alter system register;
I will continue to dig and will post any solutions that we are offered.
Regards,
Doug
On May 16, 2018 at 12:41 PM Oleksandr Denysenko <odenysenko@xxxxxxxxx> wrote:
DOUG,
they are not needed until you are using GI, without it you have to use
them for automation.
it any case, now we have situation when broker has stopped services at GI
level, but they are up at instance level and registered with listener
so, it may be:
* manual service startup
* trigger
* some king of magic
o like oracle BUG ... ;)
what to do next:
* check instance alert.log at the time of switchover. you have to see
services reset and after set again - that have to be interesting
* check listener.log at the time of switchover for services
deregistration and registration again or no such operations
* check for GI Agent log for services related activities
Best Regards,
Oleksandr Denysenko
16.05.2018 19:31, DOUG KUSHNER пишет:
> >
We have not implemented any fail-over triggers. It was my
understanding that they were required pre-11.2, before the role attribute
had been added to service config, but in 11.2.0.4 and later are not needed.
> > > On May 16, 2018 at 12:15 PM Oleksandr Denysenko
<odenysenko@xxxxxxxxx> mailto:odenysenko@xxxxxxxxx wrote:
Niall,
documentation says
https://docs.oracle.com/database/121/NETAG/advcfg.htm#sthref817 that it's
ON by default
> > > >
Connect-time failover is turned on by default forBest Regards,
multiple address lists (ADDRESS_LIST), connect descriptors
(DESCRIPTION), and multiple connect descriptors (DESCRIPTION_LIST).
> > >
Oleksandr Denysenko
16.05.2018 18:57, Niall Litchfield пишет:
> > > > If you are doing client-side
failover your TNS entry would need to be (I think)
> >
ABCR_USR =
(DESCRIPTION =
(FAILOVER=ON)
(ADDRESS_LIST =
(ADDRESS = (PROTOCOL = TCP)(HOST =
abcr-prim-scan)(PORT = 1521)) <<< This is new standby
(ADDRESS = (PROTOCOL = TCP)(HOST =
abcr-stby-scan)(PORT = 1521)) <<< This is new primary
)
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = ABCR_USR)
)
)
You should check the Net Services Administration Guide
to make sure I've got that in the right place (I likely haven't). The
principle being though that if you aren't using TAF/FAN then its up to
the client to retry addresses until one succeeds.
On Wed, May 16, 2018 at 9:55 AM, Doug Kushner <
dougk5@xxxxxxx mailto:dougk5@xxxxxxx ;> wrote:
> > > > > Additional info as
requested, all from the first node of the new standby RAC server.
NAME TYPE
VALUE
------------------------------------ -----------
-------------------------------------------
db_unique_name string
ABCR1
service_names string
ABCR_PERL,ABCR_TSAM,ABCR_USR,A BCR_WEB,ABCR1
The HOST names in the following ADDRESS_LIST are
DNS cnames that alias the scan names on each of the platforms.
ABCR_USR =
(DESCRIPTION =
(ADDRESS_LIST =
(ADDRESS = (PROTOCOL = TCP)(HOST =
abcr-prim-scan)(PORT = 1521)) <<< This is new standby
(ADDRESS = (PROTOCOL = TCP)(HOST =
abcr-stby-scan)(PORT = 1521)) <<< This is new primary
)
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = ABCR_USR)
)
)
Listener entries for the other ABCR_* services
are identical to the ABCR_USR entry and have been omitted.
LSNRCTL> services
Connecting to
(ADDRESS=(PROTOCOL=tcp)(HOST=)(PORT=1521))
Services Summary...
Service "ABCR1" has 1 instance(s).
Instance "ABCR11", status READY, has 1
handler(s) for this service...
Handler(s):
"DEDICATED" established:1504 refused:0
state:ready
LOCAL SERVER
Service "ABCR1_DGB" has 1 instance(s).
Instance "ABCR11", status READY, has 1
handler(s) for this service...
Handler(s):
"DEDICATED" established:1504 refused:0
state:ready
LOCAL SERVER
Service "ABCR1_DGMGRL" has 1 instance(s).
Instance "ABCR11", status UNKNOWN, has 1
handler(s) for this service...
Handler(s):
"DEDICATED" established:32 refused:1
LOCAL SERVER
Service "ABCR_USR" has 1 instance(s).
Instance "ABCR11", status READY, has 1
handler(s) for this service...
Handler(s):
"DEDICATED" established:1504 refused:0
state:ready
LOCAL SERVER
The command completed successfully
On 5/16/2018 12:19 AM, Oleksandr Denysenko wrote:
> > > > > >
Hello.
please, show from new standby:
* show parameter db_unique_name
* show parameter service_name
* lsnrctl services
* full desctiption of used TNS
connection string
Best Regards,
Oleksandr Denysenko
16.05.2018 8:19, Doug Kushner пишет:
> > > > > > > Hi all,
> > > > >
Please be kind as this is my first
posting to this list.
We have two RAC platforms with
12.2.0.1 GI and 11.2.0.4 RDBMS replicating with Data Guard.
Switchover testing was successful using the broker with the roles
switching successfully. Services have been configured with the
primary role on both platforms and as expected, after switchover
the services are stopped on the new standby and started on the
new primary. The applications are not TAF/FAN aware, so these
tests assume for the time being that we are shutting down the
apps (external connections) and restarting them after the
switchover.
Now for the question... We expected
as part of this switchover, that the services which were stopped
would automatically be de-registered from the listener on the new
standby side. However, they are still registered, and we are
wondering if it is our expectations or our configuration that is
at fault.
With the services still registered on
the standby side, sqlplus connection attempts to the service
result in an ORA-01033 error, since the standby address is the
first of the two addresses in the TNS alias.
Have not been able to find any info
regarding deregistration of services in the listener after a
switchover, so thought I would ask the experts.
Thanks,
Doug
--
//www.freelists.org/webpage/oracle-l ;
//www.freelists.org/webpage/oracle-l
> > > > > >
> > > >
--
Niall Litchfield
Oracle DBA
http://www.orawin.info
> > >
>