Comments

I have changed the possibility to post comments from anonymous (all) to Registered Users. Reason: loads of spam comments in my mailbox.

Event ID 110 and 117 (annoying database error)

I observe the issue described below in both AX 4.0.x and AX 2009 SP1 x solutions running SQL Server 2005 and 2008 (the version of SQL Server Native Client is the same). This is also reported by others in several Communities like microsoft.public.axapta (last one beeing a post by Mike titled "AOS errors" from February 25 2010 where one suggestion is to syncronize the SYSLASTVALUE table).

Eventlog, application

#1 Event ID 110

Object Server 01: Dialog issued for client-less session 1: Cannot edit a record in LastValue (SysLastValue). User ID: , AdminUserSetup.
The SQL database has issued an error.

#2 Event ID 117

Object Server 01:
The database reported (session 1 (-AOS-)):
[Microsoft][SQL Native Client][SQL Server]Changed language setting to us_english.
The SQL statement was:
"UPDATE SYSLASTVALUE SET VALUE=?,RECVERSION=? WHERE (((((((USERID=?) AND (RECORDTYPE=?)) AND (ELEMENTNAME=?)) AND (DESIGNNAME=?)) AND (ISKERNEL=?)) AND (COMPANY=?)) AND (RECVERSION=?))"

I don't know if this has any impact on the solutions at all and I have also found some information on Partnersource describing this as "noice". And I also find it a little bit hard to understand why a syncronization of the system table SYSLASTVALUE should be needed since the real error message in fact is stating "Changed language setting to us_english.".

The reason for posting this is that I find it a little bit hard to understand why something is logget as an error situation if this is "just noice". AX is in general bad at categorizing events in event log application, but I would like to share some experience with others regarding this. Since it's reported from the ODBC layer it could be related to settings like default language for the DBO user (aka AOS Service Account) or inproper settings for Regional Settings for the same Service Account. This is a couple of my own teories left to verify, but bottom line I find it hard to be "just noice".

Hopefully we can start a good discussion around this and bring the issue forward to the system vendor instead of reporting it individually (perhaps a bigger impact).

Update March 10 2010:
For those of you not reading the comments, I would like to point out that Chris pointed to an important KB-article from Microsoft. After looking a little bit more at the details (Googled for NativeError 5703), I found another reference on MSDN. Quote:

The SQL Server ODBC driver returns SQL_SUCCESS_WITH_INFO on a successful SQLConnect, SQLDriverConnect, or SQLBrowseConnect. When an ODBC application calls SQLError after getting SQL_SUCCESS_WITH_INFO, it can receive the following messages:

5701—indicates SQL Server initially putting the user's context into the default database defined at the server for the login ID used in the connection
5703—indicates the language being used on the server

If either the ODBC data source has a default database specified or the application specified the DATABASE keyword on SQLDriverConnect or SQLBrowseConnect, there will be a second 5701 message that indicates the user's context has been switched to the database requested.

My conclusion is that the AOS kernel treat every error raised from the ODBC layer as an error. If Microsoft could implement a change to filter out NativeError 5701 and 5703, a lot of confusion could be saved and not at least a lot of time investigating this issue could also be saved.

Until further notice, this is my conclusion.

Again, thank you Chris for leaving this valuable information.

Update #2 March 11 2010:

I have today posted this as a suggestion on the MS Dynamics AX Suggestions Home page at Microsoft Connect. Please sign in at Connect and vote this up if you find this issue to be a valuable improvement.

Update #3 May 11 2010:

The EMEA Dynamics AX Support team recently released a post giving another explanation. It can be found here.

AX Load Balancing

A tree tiered architecture has been around since AX 4.0. In AX 2009 you have 3 possible AOS configurations:

1. Single AOS instance(s)
2. Load Balanced AOS instances without a dedicated load balancer
3. Load Balanced AOS instances with a dedicted load balancer

After experimenting a little bit with all of the possible configurations, I find it a little bit difficult to separate configuration 2 and 3 above. The only difference is the presence of one (or several) AOS instance(s) defined as dedicated load balancers only handling the load balancing functionality (no end user sessions) for AX clients that whish to log on to AX (handshake). So the big question is when to use a configuration with a dedicated load balancer. I can't see any reason or benefit other than limiting the need to adjust the client configurations when AOS instances move out and in of the cluster. The only benefit deploying a dedicated load balancer is that this will be the only AOS instance referenced by the clients and that this makes the client configuration more static compared to configuration 2 when every AOS instance in fact both is acting as a load balancer and handling client sessions.

And information from Microsoft states that a dedicated load balancer only is needed when you have more than 4 AOS instances in the same cluster.

All combinations in the client configurations will work (no difference wheter you only reference the dedicated load balancing instance and/or all AOS instances).

Any comments around this matter will be appreciated.

Short note on AX and database mirroring in SQL Server 2008

The last days, I have been working with a customer running AX 2009 and SQL Server 2008 SP1 Standard (x64) on Windows Server 2008 SP2. The solution suffered from general performance issues that possibly could have a lot of sources. As usual my attention was around SQL Server and this time I started looking at Wait Stats which showed a waste amount of waits related to mirroring. After some Googling, I was a little bit confused about this beeing normal or not. The nature of database mirroring could in fact result in high Wait Stat values since the processes involved, is mostly sleeping (suspended). But I choosed to consider this as a possible source since the nature of database mirroring in my oppinion, does'nt fit the nature of AX as a classic OLTP application. And since disabling database mirroring is an isolated and low risk operation, we choosed to stop mirroring for all database (around 10). After this, the overall performance increased and the customer reported this almost immediately. It's a little bit early to conclude, but the initial responce seems pretty promising.

Minor observation installing AX 2009

Some time back, I did a QA run on one installation suffering some performance issues. The consultant originally installing AX choosed to use the installer to create the database (shown in the picture below). Personally I prefer to create the database manually, both to get control over every aspect and it usually don't take more than 30 minutes to get everything in place manually. The solution at hand was running SQL Server 2008.

While looking into the database configuration (sp_helpdb), I noticed that the compatibility level was set to 90 which corresponds to SQL Server 2005. At this stage, I asked the consultant doing the installation why he had choosen to set the compatibility level to 90 and he then answered that he always used the AX installer to create the database.

Without testing this option when installing AX, it seems like Microsoft has provided a template database beeing attached to the specified SQL Server instance by the installer in the security context of the setup user. I guess the installer asks for some information like collation, file locations etc. but compatibility level keeps untouched.

Compatibility level is a database level configuration option and it is used by the database instance to decide which functionality the database instance should "activate" for each database. It's mostly used to provide backwards compatibility during migration of databases, but it could also impact the ability to utilize new features.

A good reference can be found at MSDN and compatibility level could possibly be a source for some time consuming investigations. It seems like compatibility level = 90 is OK even for an AX database running on SQL Server 2008, but I don't see any reason for not setting it to 100 to make sure all new features in SQL Server 2008 is enabled for the AX database.

Or always create the database manually using the system database model as the template (will in most situations match the version of SQL Server).

A new year approaching

With another week to go and the Cristmas hollidays coming, it's time to reflect on the past and coming year. I think 2009 was a good year for AX 2009 despite the product beeing released in 2008. The main reason for 2009 beeing a good year for AX 2009 is mainly based on the Gartner report positioning AX 2009 the number one (#1) leader in the magic quadrant for what we call SMB here in Norway. This could of course be marketing hoopla, but if you read the report, Gartner pays special attention to "Microsoft delivering on their vision". Personally I find this interesting and I suggest that every partner study the current Roadmap and Statement of Direction (SOD) to see what Microsoft is planning (visionary) for AX in the future. In my mind, AX 2009 did'nt bring a lot of news on the techology side (the new batch framework and the support for UTC is perhaps the most important improvements), but overall MS managed to position AX as a real challenger and maybe also a winner in the combat against SAP and Oracle/PeopleSoft.

Based on the number of hotfixes released for AX 2009 and SP1, MS still has some challenges regarding quality. Maybe a slowdown in the release cycle would be a good advise and maybe also broaden the early adoption program to gain some more experience from the field. Again this is really not special to AX, since we have to remember that all software is manufactured by humans and no human is free of errors. All in all I think AX 2009 was a great step forward at least with regards to technology and architecture, since the product is well positioned to compete with the biggest rivals even in the upper, right quadrant. Finally "MS got rid of most of (every?) not so industry standard implementation as expected" and lifted the product up to the MS level of integration (still some ground to cover, but yet greatly improved).

So it's exiting times generally and with regards to AX. I'm positive about the year to come and also the next decade. From the Roadmap and SOD, I read a steady growth both vertically and horizontaly with regards to functionality. We are already seeing some evidence of this with MS buying verticals from partners (like the POS vertical from a Danish world wide partner financing? a rather special twist in buing a highly specialized ERP solution called Guideix A/S). The future story of this evolution of one of my previous employers, will be very interesting to follow. Driven by market, customers or product/technology? Who knows...

Anyway I'm optimistic about the future of AX and I seriously meen that AX 2009 brought AX closer (really close) to what the product has been marketed as since the beginning (Damgaard back in 1998 - more than 10 years ago). At the same time I have to admire the Damgaard brothers for the introduction of a true 3-tired solution (was it 2.5?) that MS now has adopted and brought forward to a pure 3-tiered solution with no other choices. Add the rather annoying AOCP beeing replaced with MS RPC in AX 4.0 and you get the picture.

One final message to all x++ developers out there, is to take the time to study and understand the difference between the different caching schemes available in AX (look at the DEV III course documentation or Inside AX 4.0/2009 book). This is key knowledge to utilize the 3-tier architecture in AX 4.0 and 2009 (also valid for earlier versions when running in 3-tire thin configurations) and at the same time, rather complicated. From my experience, this is an area that get far to less attention and that has a big potential for optimizations (and performance gains).

Without digging deeper into the philosopical area, I would take the opportunity to wish my few (but valuable) readers, a happyy Christmas and an interesting new AX year (and decade)!

See you in 2010! So long...

DAS or SAN for SQL Server part II

I thought a follow up was adequate for this one. After looking for some more views, oppinions and recommendations, I found the following posts (3 part) on the blog for the Microsoft SQL Server Development Customer Advisory Team:

Deploying SQL Server with SAN #1
Deploying SQL Server with SAN #2
Deploying SQL Server with SAN #3