Replication error at the end of the completion of the snapshot agent: ConnectionRead (recv())

Replication error at the end of the completion of the snapshot agent: ConnectionRead (recv())

Post by Erd?l BIRAME » Sat, 02 Oct 1999 04:00:00



I have set up a replication with the following topology.
NT Server Enterprise Edition + SP5 + SQL Server 7.0 Enterprise Edition + SP1
Publisher, distributer and subscribers are all member of the same domain.
Publisher is in an other city
Distributer and subscriber computers are on my desk
SQL-Server and SQL-Agent services on all three computers use the same
NT-service-account, which is a domain user account with administrative
privilieges
Replication agents log on the computers impersonating the SQLAgent service
account
Publisher logs an the Distributor using a custom password for the
distributor_admin built-in account.
When I run the replication with a small publication let's say with only a
single table, creation of the initial snapshot works fine
When I run the replication with the actual publication with 22 tables
comprising 1,3 GB of data, it takes 130 minutes, the schema files are
created correctly, the bcp-files are also created correctly. But after
everything is done correctly, in the final step (i.e. the final entry in the
history of the snapshot agent) the following error is reported:
Publisher (DESCLS01), Distributor(DEV05), Subscriber(DEV06), PublicationDB
(GNMMain), Publication (GNMMain), DistributionDB (GNMMainDistribution)
Publisher:        DESCLS01
Agent name:    DESCLS01-GNMMain-GNMMain-4
Publication:    GNMMain
Last action:     ConnectionRead (recv()).
Category:        Data Source
Source:            DESCLS01
Number:           0
Full error message:    ConnectionRead (recv()).

The SQLAgent error log and SQL-Server log don't contain any records on this
error.

The NT event log has the following entry:

Date:                Date of the error
Time:                Time of the error
User:                N/A
Computer:        Computer name of the distributor
Event ID:           203
Source:            SQLServerAgent
Type:                Information
Category:          Job Engine
Description:
SubSystem Message - Job 'DESCLS01-GNMMain-GNMMain-4'
(0xC86FF23E0D77D311ACE700508B0C276F), step 2 - ConnectionRead (recv()).

Explaination:
Job 'DESCLS01-GNMMain-GNMMain-4'  is the name of the SQLAgent-Job
representing the snapshot-agent.

Step2 which is causing the error is:

Step Name:    Run agent
Type:              Replication snapshot
Command:       -Publisher [DESCLS01] -PublisherDB [GNMMain] -Distributor
[DEV05] -Publication [GNMMain] -DistributorSecurityMode 1

I would be endlessly thankful for any help.

Erd?l BIRAMEN

 
 
 

Replication error at the end of the completion of the snapshot agent: ConnectionRead (recv())

Post by Graeme Slogrov » Sat, 02 Oct 1999 04:00:00



Quote:> I have set up a replication with the following topology.
> NT Server Enterprise Edition + SP5 + SQL Server 7.0 Enterprise Edition +
SP1
> Publisher, distributer and subscribers are all member of the same domain.
> Publisher is in an other city

What speed link do you have between the cities ?

I found that creating a custom agent property for the replication process,
with very small packet sizes, helped me resolve problems with timeouts
during the replication.

 
 
 

Replication error at the end of the completion of the snapshot agent: ConnectionRead (recv())

Post by eddi » Fri, 08 Oct 1999 04:00:00


The only thing coming to my mind is that I had a problem replicating a
large table because after the data was BCP'ed over from publisher to
subscriber, I would keep getting timeouts that mystified me until I
realized that the subscriber was taking too long creating the 15 or so
INDEXES on the large table. After that, I told replication NOT to
create all the non-clustered indexes during the snapshot. I scripted
them and ran that later, after the table replicated successfully.

Ed R.

On Fri, 1 Oct 1999 15:08:22 +0200, "Erd?l BIRAMEN"


>I have set up a replication with the following topology.
>NT Server Enterprise Edition + SP5 + SQL Server 7.0 Enterprise Edition + SP1
>Publisher, distributer and subscribers are all member of the same domain.
>Publisher is in an other city
>Distributer and subscriber computers are on my desk
>SQL-Server and SQL-Agent services on all three computers use the same
>NT-service-account, which is a domain user account with administrative
>privilieges
>Replication agents log on the computers impersonating the SQLAgent service
>account
>Publisher logs an the Distributor using a custom password for the
>distributor_admin built-in account.
>When I run the replication with a small publication let's say with only a
>single table, creation of the initial snapshot works fine
>When I run the replication with the actual publication with 22 tables
>comprising 1,3 GB of data, it takes 130 minutes, the schema files are
>created correctly, the bcp-files are also created correctly. But after
>everything is done correctly, in the final step (i.e. the final entry in the
>history of the snapshot agent) the following error is reported:
>Publisher (DESCLS01), Distributor(DEV05), Subscriber(DEV06), PublicationDB
>(GNMMain), Publication (GNMMain), DistributionDB (GNMMainDistribution)
>Publisher:        DESCLS01
>Agent name:    DESCLS01-GNMMain-GNMMain-4
>Publication:    GNMMain
>Last action:     ConnectionRead (recv()).
>Category:        Data Source
>Source:            DESCLS01
>Number:           0
>Full error message:    ConnectionRead (recv()).

>The SQLAgent error log and SQL-Server log don't contain any records on this
>error.

>The NT event log has the following entry:

>Date:                Date of the error
>Time:                Time of the error
>User:                N/A
>Computer:        Computer name of the distributor
>Event ID:           203
>Source:            SQLServerAgent
>Type:                Information
>Category:          Job Engine
>Description:
>SubSystem Message - Job 'DESCLS01-GNMMain-GNMMain-4'
>(0xC86FF23E0D77D311ACE700508B0C276F), step 2 - ConnectionRead (recv()).

>Explaination:
>Job 'DESCLS01-GNMMain-GNMMain-4'  is the name of the SQLAgent-Job
>representing the snapshot-agent.

>Step2 which is causing the error is:

>Step Name:    Run agent
>Type:              Replication snapshot
>Command:       -Publisher [DESCLS01] -PublisherDB [GNMMain] -Distributor
>[DEV05] -Publication [GNMMain] -DistributorSecurityMode 1

>I would be endlessly thankful for any help.

>Erd?l BIRAMEN

 
 
 

Replication error at the end of the completion of the snapshot agent: ConnectionRead (recv())

Post by Erd?l BIRAME » Sat, 09 Oct 1999 04:00:00


Thany you for your help Eddie.

I have already found out that extensive index creation (appr. 1 hour) was
causing the error and I also know that I could turn out the replication of
indexes feature to solve this issue. But the problem is, that I need this
feature, as there are frequent schema changes occurring and the guys in this
firm have absolutely no time and willingness to care about replication once
I have set it up and left here. (I am just an independent consultant hired
to set up the replication)

So what I need is to prolong the timeout period. What would you suggest ?

The QueryTimeOut parameter of the agents     OR
The inactivity threshold in the 'Refresh rate and settings' dialog ?

And also have you ever experimented with the various agent parameters? Did
you find out any good parameter settings for replication of large tables
(250-800 MB) over slow VPN?
Especially BcpBatchSize, CommitBatchSize, CommitBatchThreshold,
MaxBcpThreads, MaxDeliveredTransactions

And did you also find out that the distribution agent has an undocumented
parameter called SkipFailureLevel in the 'Replication agent profile details'
dialog?

Erd?l BIRAMEN


> The only thing coming to my mind is that I had a problem replicating a
> large table because after the data was BCP'ed over from publisher to
> subscriber, I would keep getting timeouts that mystified me until I
> realized that the subscriber was taking too long creating the 15 or so
> INDEXES on the large table. After that, I told replication NOT to
> create all the non-clustered indexes during the snapshot. I scripted
> them and ran that later, after the table replicated successfully.

> Ed R.

> On Fri, 1 Oct 1999 15:08:22 +0200, "Erd?l BIRAMEN"

> >I have set up a replication with the following topology.
> >NT Server Enterprise Edition + SP5 + SQL Server 7.0 Enterprise Edition +
SP1
> >Publisher, distributer and subscribers are all member of the same domain.
> >Publisher is in an other city
> >Distributer and subscriber computers are on my desk
> >SQL-Server and SQL-Agent services on all three computers use the same
> >NT-service-account, which is a domain user account with administrative
> >privilieges
> >Replication agents log on the computers impersonating the SQLAgent
service
> >account
> >Publisher logs an the Distributor using a custom password for the
> >distributor_admin built-in account.
> >When I run the replication with a small publication let's say with only a
> >single table, creation of the initial snapshot works fine
> >When I run the replication with the actual publication with 22 tables
> >comprising 1,3 GB of data, it takes 130 minutes, the schema files are
> >created correctly, the bcp-files are also created correctly. But after
> >everything is done correctly, in the final step (i.e. the final entry in
the
> >history of the snapshot agent) the following error is reported:
> >Publisher (DESCLS01), Distributor(DEV05), Subscriber(DEV06),
PublicationDB
> >(GNMMain), Publication (GNMMain), DistributionDB (GNMMainDistribution)
> >Publisher:        DESCLS01
> >Agent name:    DESCLS01-GNMMain-GNMMain-4
> >Publication:    GNMMain
> >Last action:     ConnectionRead (recv()).
> >Category:        Data Source
> >Source:            DESCLS01
> >Number:           0
> >Full error message:    ConnectionRead (recv()).

> >The SQLAgent error log and SQL-Server log don't contain any records on
this
> >error.

> >The NT event log has the following entry:

> >Date:                Date of the error
> >Time:                Time of the error
> >User:                N/A
> >Computer:        Computer name of the distributor
> >Event ID:           203
> >Source:            SQLServerAgent
> >Type:                Information
> >Category:          Job Engine
> >Description:
> >SubSystem Message - Job 'DESCLS01-GNMMain-GNMMain-4'
> >(0xC86FF23E0D77D311ACE700508B0C276F), step 2 - ConnectionRead (recv()).

> >Explaination:
> >Job 'DESCLS01-GNMMain-GNMMain-4'  is the name of the SQLAgent-Job
> >representing the snapshot-agent.

> >Step2 which is causing the error is:

> >Step Name:    Run agent
> >Type:              Replication snapshot
> >Command:       -Publisher [DESCLS01] -PublisherDB [GNMMain] -Distributor
> >[DEV05] -Publication [GNMMain] -DistributorSecurityMode 1

> >I would be endlessly thankful for any help.

> >Erd?l BIRAMEN

 
 
 

Replication error at the end of the completion of the snapshot agent: ConnectionRead (recv())

Post by eddi » Sat, 09 Oct 1999 04:00:00


I have not played with many of the switches.  I set the QueryTimeOut
period way up there and hoped for the best.

On Fri, 8 Oct 1999 12:58:20 +0200, "Erd?l BIRAMEN"


>Thany you for your help Eddie.

>I have already found out that extensive index creation (appr. 1 hour) was
>causing the error and I also know that I could turn out the replication of
>indexes feature to solve this issue. But the problem is, that I need this
>feature, as there are frequent schema changes occurring and the guys in this
>firm have absolutely no time and willingness to care about replication once
>I have set it up and left here. (I am just an independent consultant hired
>to set up the replication)

>So what I need is to prolong the timeout period. What would you suggest ?

>The QueryTimeOut parameter of the agents     OR
>The inactivity threshold in the 'Refresh rate and settings' dialog ?

>And also have you ever experimented with the various agent parameters? Did
>you find out any good parameter settings for replication of large tables
>(250-800 MB) over slow VPN?
>Especially BcpBatchSize, CommitBatchSize, CommitBatchThreshold,
>MaxBcpThreads, MaxDeliveredTransactions

>And did you also find out that the distribution agent has an undocumented
>parameter called SkipFailureLevel in the 'Replication agent profile details'
>dialog?

>Erd?l BIRAMEN



>> The only thing coming to my mind is that I had a problem replicating a
>> large table because after the data was BCP'ed over from publisher to
>> subscriber, I would keep getting timeouts that mystified me until I
>> realized that the subscriber was taking too long creating the 15 or so
>> INDEXES on the large table. After that, I told replication NOT to
>> create all the non-clustered indexes during the snapshot. I scripted
>> them and ran that later, after the table replicated successfully.

>> Ed R.

>> On Fri, 1 Oct 1999 15:08:22 +0200, "Erd?l BIRAMEN"

>> >I have set up a replication with the following topology.
>> >NT Server Enterprise Edition + SP5 + SQL Server 7.0 Enterprise Edition +
>SP1
>> >Publisher, distributer and subscribers are all member of the same domain.
>> >Publisher is in an other city
>> >Distributer and subscriber computers are on my desk
>> >SQL-Server and SQL-Agent services on all three computers use the same
>> >NT-service-account, which is a domain user account with administrative
>> >privilieges
>> >Replication agents log on the computers impersonating the SQLAgent
>service
>> >account
>> >Publisher logs an the Distributor using a custom password for the
>> >distributor_admin built-in account.
>> >When I run the replication with a small publication let's say with only a
>> >single table, creation of the initial snapshot works fine
>> >When I run the replication with the actual publication with 22 tables
>> >comprising 1,3 GB of data, it takes 130 minutes, the schema files are
>> >created correctly, the bcp-files are also created correctly. But after
>> >everything is done correctly, in the final step (i.e. the final entry in
>the
>> >history of the snapshot agent) the following error is reported:
>> >Publisher (DESCLS01), Distributor(DEV05), Subscriber(DEV06),
>PublicationDB
>> >(GNMMain), Publication (GNMMain), DistributionDB (GNMMainDistribution)
>> >Publisher:        DESCLS01
>> >Agent name:    DESCLS01-GNMMain-GNMMain-4
>> >Publication:    GNMMain
>> >Last action:     ConnectionRead (recv()).
>> >Category:        Data Source
>> >Source:            DESCLS01
>> >Number:           0
>> >Full error message:    ConnectionRead (recv()).

>> >The SQLAgent error log and SQL-Server log don't contain any records on
>this
>> >error.

>> >The NT event log has the following entry:

>> >Date:                Date of the error
>> >Time:                Time of the error
>> >User:                N/A
>> >Computer:        Computer name of the distributor
>> >Event ID:           203
>> >Source:            SQLServerAgent
>> >Type:                Information
>> >Category:          Job Engine
>> >Description:
>> >SubSystem Message - Job 'DESCLS01-GNMMain-GNMMain-4'
>> >(0xC86FF23E0D77D311ACE700508B0C276F), step 2 - ConnectionRead (recv()).

>> >Explaination:
>> >Job 'DESCLS01-GNMMain-GNMMain-4'  is the name of the SQLAgent-Job
>> >representing the snapshot-agent.

>> >Step2 which is causing the error is:

>> >Step Name:    Run agent
>> >Type:              Replication snapshot
>> >Command:       -Publisher [DESCLS01] -PublisherDB [GNMMain] -Distributor
>> >[DEV05] -Publication [GNMMain] -DistributorSecurityMode 1

>> >I would be endlessly thankful for any help.

>> >Erd?l BIRAMEN