I agree with Nassyam, the errors clearly point to network problems between the
production and standby sites. I have had the same on various occasions,
especially when production traffic and log shipping used the same network
devices. Congestion can take place on various levels in the OSI layers of the
servers involved, not just the network devices.
The solution is not to increase the file size, as the network traffic on copy
only increases, but enabling real-time apply as suggested. When real-time apply
is enabled, the production log writer will write transactions to the standby
log directly, in tandem with the local redo log. When the log is archived, a
complete copy already exists on the DG server and no further network copy is
needed. In addition, changes are applied to the standby database as they occur,
so activity on that server is steady and does not rise and fall as would be the
case with log shipping. This does not increase the load on the production
server, but you must take care that the production database will not hang if
for any reason the log writer cannot write to the standby log.
It can also not do any harm to check that the network configuration is sane,
e.g. that a ping between the two servers does not show any lost packets.
Ususally when this happens, a server is set up to use two or more NICs in
tandem (i.e. teamed/bonded/etc.) but the switches they are connected to are not
set up to support this. E.g. with 2 teamed/bonded/.. NICs, you'd see about 50%
packet loss if the switches are not configured correctly.
Cheers,
Tony
On 09/01/16 12:49, Upendra nerilla wrote:
I reviewed the network throughput and we aren't saturated from the network pipe
perspective. We do have riverbed network gear which is compressing and
optimizing the archive transport. I am trying to determine where the bottleneck
is..
Since the logs aren't getting shipped to the DR site, I wanted to get that
resolved.
BTW, I increased redolog size in Prod from 1G to 2G, I still see missing
archivelogs in the DR site.
Real-time apply is not really a requirement, I thought it has more overhead to
the system compared to archive shipping.. With the amount of data I am pushing,
would it help?
I checked the SDU size, we are using the default. How should I determine if SDU
needs to be increased?
I haven't done anything with chunk size, any suggestions on where should I
start?
Thanks
-Upendra
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Subject: Re: Troubleshooting log shipping
From: nassyambasha@xxxxxxxxx
Date: Sat, 9 Jan 2016 05:44:08 +0530
CC: oracle-l@xxxxxxxxxxxxx
To: nupendra@xxxxxxxxxxx
Okay this information clears me Data Guard have challenge from network
speed/bandwidth.
One simple question, why not Real-Time Apply ? No need to depend until the
archive log archived and transmitted.
If it is already implemented then, you must consider optimizing the network by
SDU size with large chunk.
Sent from my iPhone
On 08-Jan-2016, at 11:47 PM, Upendra nerilla <nupendra@xxxxxxxxxxx
<mailto:nupendra@xxxxxxxxxxx>> wrote:
Hi Nassyam et al -
Thanks for your suggestion in setting the archive_dest_state to defer and
enable. That cleared the error. I don't see any errors now.
From looking at the v$archived_log I see that we are behind by a few
hundred logs..
When I check those files with your second query, I do see that a bunch of
archivelogs haven't reached standby.
A couple of followup questions:
1. This is a busy system, redolog size is 1G and in peak we generate about
100-120 logs in an hour. Is it possible that archiver gets too busy and can't
handle this.
Is there a way to determine that? Should I consider increasing the redo log
size to handle this?
2. Does anyone have a script or suggestions on a way to automate copying
the missing archive logs to the Data guard site? Oracle is automatically
pulling these files at the time of logapply. But because of the
latency/throughput between the sites, I would like to copy them ahead of time
so the logapply is going through faster.
Thanks
-Upendra
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
From: nassyambasha@xxxxxxxxx <mailto:nassyambasha@xxxxxxxxx>
Date: Fri, 8 Jan 2016 19:49:33 +0530
Subject: Re: Troubleshooting log shipping
To: nupendra@xxxxxxxxxxx <mailto:nupendra@xxxxxxxxxxx>
CC: oracle-l@xxxxxxxxxxxxx <mailto:oracle-l@xxxxxxxxxxxxx>
Hello Upendra,
How about the network between primary and standby site? For the best
practice you can do like this to refresh.
Primary> alter system set log_archive_dest_state_3='defer';
Primary> alter system set log_archive_dest_state_3='enable';
Primary> alter system archive log current;
Then check again the above query for the latest errors and also standby
alert log if any sequences are catching up.
From the alert log those errors, i have seen many times the situation can be vary
like network issue or unable to communicate with the background processes Arch -->
(LNS) --> RFS.
Provide the below information first to understand what archive gap you have.
A) Primary> select thread#,max(sequence#) from v$archived_log group by
thread#;
B) Standby>select thread#,max(sequence#) from v$archived_log where
applied='YES' group by thread#;
If there is any GAP on standby database then we need to check if the
archives whether reached to standby or not.
standby> select thread#,sequence#,name, applied from v$archived_log where
sequence#=&outoutBThread1seq and thread#=1;
standby> select thread#,sequence#,name, applied from v$archived_log where
sequence#=&outoutBThread2seq and thread#=2;
If they are unavailable on standby then we have to check those missing
archives are available on primary using the same above queries. if available
then troubleshooting is required why unable to ship to standby. If archives
unavailable then you know what to do :)
Thanks.
On Fri, Jan 8, 2016 at 7:24 PM, Upendra nerilla <nupendra@xxxxxxxxxxx
<mailto:nupendra@xxxxxxxxxxx>> wrote:
oops.. sorry about that.. still waking up.. :)
Here is the output from the second query:
SEVERITY ERROR_CODE timestamp MESSAGE
--------------- ---------- --------------------
----------------------------------------------------------------------
Error 16198 08-JAN-2016 07:23:37 WARN: ARC3: Terminating
pid 18564 hung on an I/O operation
Error 16198 08-JAN-2016 07:23:39 ARC3: Error 16198 due
to hung I/O operation to LOG_ARCHIVE_DEST_3
I see the following alert.log entry in Prod:
Thu Jan 07 19:09:15 2016
LNS: Standby redo logfile selected for thread 1 sequence 435322 for
destination LOG_ARCHIVE_DEST_2
krsv_proc_kill: Killing 1 processes (Process by index)
krsv_proc_kill: Killing 1 processes (Process by index)
ARC1: Error 16198 due to hung I/O operation to LOG_ARCHIVE_DEST_3
ARC1: Detected ARCH process failure
ARC1: Detected ARCH process failure
ARC1: STARTING ARCH PROCESSES
Thanks
-Upendra
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
From: nassyambasha@xxxxxxxxx <mailto:nassyambasha@xxxxxxxxx>
Date: Fri, 8 Jan 2016 18:50:04 +0530
Subject: Re: Troubleshooting log shipping
To: nupendra@xxxxxxxxxxx <mailto:nupendra@xxxxxxxxxxx>
CC: oracle-l@xxxxxxxxxxxxx <mailto:oracle-l@xxxxxxxxxxxxx>
Hello Upendra,
I understand. But you missed to provide the first query i.e.
set line 120 pages 100
col severity for a15
col message for a70
col timestamp for a20
select severity,error_code,to_char(timestamp,'DD-MON-YYYY HH24:MI:SS')
"timestamp" , message from v$dataguard_status where dest_id=3;
Thanks.
On Fri, Jan 8, 2016 at 6:45 PM, Upendra nerilla <nupendra@xxxxxxxxxxx
<mailto:nupendra@xxxxxxxxxxx>> wrote:
Hi Nassyam,
I believe the configuration is correct, otherwise I wouldn't be receiving any
archivelogs at the destination. The DB_unique_name is set to "SID_DR" in the
data guard site and there is a corresponding TNS entry in the PROD site (with the same
name) pointing to the data guard site.
The issue is I am receiving SOME logs but not ALL the archivelogs
at the data guard site.
As I pasted the output of the archivelogs in my earlier message,
see the sequence numbers are all over the place..
This is from ASM:
512 1860501 952576512 954204160
thread_1_seq_*434986*.5778.900455621
512 1883168 964182016 965738496
thread_1_seq_*434991*.8962.900456295
512 1850346 947377152 948961280
thread_1_seq_*434997*.9060.900456899
512 1865067 954914304 956301312
thread_1_seq_*435003*.3202.900457539
512 134582 68905984 70254592
thread_1_seq_*435008*.1471.900457605
512 1848325 946342400 947912704
thread_1_seq_*435009*.3990.900458259
512 1863147 953931264 955252736
thread_1_seq_*435017*.6484.900458793
512 1834310 939166720 940572672
thread_1_seq_*435023*.8233.900459079
512 1438234 736375808 738197504
thread_1_seq_*435024*.11250.900460605
When the logapply hits the point where the archivelog is missing it
is copying from Prod, which is slowing the log apply significantly.
From alert.log at data guard:
Media Recovery Waiting for thread 1 sequence 434987 (in transit)
Here is the output of the query you sent (taken from Prod):
ID STATUS DB_MODE TYPE RECOVERY_MODE
PROTECTION_MODE SRLs ACTIVE
---------- --------- --------------- ----------
----------------------- -------------------- ---------- ----------
ARCHIVED_SEQ#
-------------
1 VALID OPEN ARCH IDLE MAXIMUM PERFORMANCE
0 0
355351
2 VALID MOUNTED-STANDBY LGWR MANAGED MAXIMUM PERFORMANCE
14 3
435651
3 VALID MOUNTED-STANDBY LGWR MANAGED MAXIMUM PERFORMANCE
14 3
353763
Thanks
-Upendra
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
From: nassyambasha@xxxxxxxxx <mailto:nassyambasha@xxxxxxxxx>
Date: Fri, 8 Jan 2016 15:08:10 +0530
Subject: Re: Troubleshooting log shipping
To: nupendra@xxxxxxxxxxx <mailto:nupendra@xxxxxxxxxxx>
CC: oracle-l@xxxxxxxxxxxxx <mailto:oracle-l@xxxxxxxxxxxxx>
Hello Upendra,
If the configuration is perfect then the redo/archive should
receive on remote destinations.
If you see below, you have mentioned service=SID_DR, can you
confirm you have used the Oracle net service which points to the related
standby database? and what about db_unique_name also you have mentioned SID_DR,
it should be the unique name but not SID, so it depends on configuration
however what the db_unique_name,sid, service configured.
"*service="SID_DR", LGWR ASYNC NOAFFIRM delay=0 optional
compression=disable max_failure=0 max_connections=1 reopen=15*
*db_unique_name="SID_DR" net_timeout=30,
valid_for=(all_logfiles,primary_role)*"
Can you send output of this below query? Run them from primary.
set line 120 pages 100
col severity for a15
col message for a70
col timestamp for a20
select severity,error_code,to_char(timestamp,'DD-MON-YYYY HH24:MI:SS')
"timestamp" , message from v$dataguard_status where dest_id=3;
select ds.dest_id id
, ad.status
, ds.database_mode db_mode
, ad.archiver type
, ds.recovery_mode
, ds.protection_mode
, ds.standby_logfile_count "SRLs"
, ds.standby_logfile_active active
, ds.archived_seq#
from v$archive_dest_status ds
, v$archive_dest ad
where ds.dest_id = ad.dest_id
and ad.status != 'INACTIVE'
order by
ds.dest_id;
Thanks.
On Fri, Jan 8, 2016 at 1:22 PM, Upendra nerilla <nupendra@xxxxxxxxxxx
<mailto:nupendra@xxxxxxxxxxx>> wrote:
Hello everyone -
Happy new year!
Oracle RAC 11.2.0.3
OEL 5.x
I have an environment in which the primary is shipping logs to
a local standby and a remote dataguard database.
What I am seeing is that, somehow not all the archivelogs are
being shipped to the remote site. At the time of log apply these logs are being
transmitted over.
Here is an example on how the redologs appear in the remote
site:
512 1860501 952576512 954204160
thread_1_seq_434986.5778.900455621
512 1883168 964182016 965738496
thread_1_seq_434991.8962.900456295
512 1850346 947377152 948961280
thread_1_seq_434997.9060.900456899
512 1865067 954914304 956301312
thread_1_seq_435003.3202.900457539
512 134582 68905984 70254592
thread_1_seq_435008.1471.900457605
512 1848325 946342400 947912704
thread_1_seq_435009.3990.900458259
512 1863147 953931264 955252736
thread_1_seq_435017.6484.900458793
512 1834310 939166720 940572672
thread_1_seq_435023.8233.900459079
512 1438234 736375808 738197504
thread_1_seq_435024.11250.900460605
Similar gaps for thread_2 as well..
Here is the configuration of the log_archive_dest_3 on primary
for the remote site:
*service="SID_DR", LGWR ASYNC NOAFFIRM delay=0 optional
compression=disable max_failure=0 max_connections=1 reopen=15
db_unique_name="SID_DR" net_timeout=30,
valid_for=(all_logfiles,primary_role)*
I do not see any errors for the redo transport in the alert log
on the primary site or on the remote DG site.
Any thoughts on how should I troubleshoot this? Are there any
known bugs around this?
Your help is appreciated.
Thanks
-Upendra
-- Nassyam Basha.
*Oracle Database Consultant*| *Pythian * <http://www.pythian.com/>
ACED Profile
<https://apex.oracle.com/pls/apex/f?p=19297:4:::NO:4:P4_ID:13140>
Oracle 11g Certified Master
<http://education.oracle.com/education/otn/NassyamBasha.htm>
Co-Author: Oracle Data Guard 11gR2
<http://www.amazon.in/Oracle-Guard-11gR2-Administration-Beginners/dp/1849687900>
Co-founder of Oraworld-team <http://www.oraworld-team.com>
Facebook <https://www.facebook.com/nassyambasha> Twitter
<https://twitter.com/oracle_ckpt> LinkedIn <https://in.linkedin.com/in/nassyambasha> Google +
<https://plus.google.com/+NassyamBasha> CKPT Blog <http://www.oracle-ckpt.com/>
-- Nassyam Basha.
*Oracle Database Consultant*| *Pythian * <http://www.pythian.com/>
ACED Profile
<https://apex.oracle.com/pls/apex/f?p=19297:4:::NO:4:P4_ID:13140>
Oracle 11g Certified Master
<http://education.oracle.com/education/otn/NassyamBasha.htm>
Co-Author: Oracle Data Guard 11gR2
<http://www.amazon.in/Oracle-Guard-11gR2-Administration-Beginners/dp/1849687900>
Co-founder of Oraworld-team <http://www.oraworld-team.com>
Facebook <https://www.facebook.com/nassyambasha> Twitter
<https://twitter.com/oracle_ckpt> LinkedIn <https://in.linkedin.com/in/nassyambasha> Google +
<https://plus.google.com/+NassyamBasha> CKPT Blog <http://www.oracle-ckpt.com/>
-- Nassyam Basha.
*Oracle Database Consultant*| *Pythian * <http://www.pythian.com/>
ACED Profile
<https://apex.oracle.com/pls/apex/f?p=19297:4:::NO:4:P4_ID:13140>
Oracle 11g Certified Master
<http://education.oracle.com/education/otn/NassyamBasha.htm>
Co-Author: Oracle Data Guard 11gR2
<http://www.amazon.in/Oracle-Guard-11gR2-Administration-Beginners/dp/1849687900>
Co-founder of Oraworld-team <http://www.oraworld-team.com>
Facebook <https://www.facebook.com/nassyambasha> Twitter <https://twitter.com/oracle_ckpt>
LinkedIn <https://in.linkedin.com/in/nassyambasha> Google +
<https://plus.google.com/+NassyamBasha> CKPT Blog <http://www.oracle-ckpt.com/>
