Hello everyone.
The ultimate question is: Do you have to use sequelink to connect TN to WM? If you don’t, what can you use instead?
We currently have TN 4.5.1 using Sequelink 5.3 to connect to WM 4.6 with an Oracle DB 8.1.7.4. Any advice would be greatly appreciated.
TIA
Dave
Dave,
For TN pre6.0 versions ie.(4.0.4.5,4.6)you have to use the sequelink to connect an external DB.(Internally TN comes with CloudScape DB for this no need of using sequellink)
HTH,
Dave,
If you pester your wM support people enough, they might give you a JDBC driver to use that will eliminate the SequeLink server requirement. That is, you supply a good sob story about SequeLink server dying, unstable, slow, what not… 8)
Hint: Look at how TN 6.* is connecting to RDBMS…
We experienced problems using the Merant proxy with IS/TN 4.6. WM supplied a driver – sl53_cj32(build38).jar – to connect TN directly to Oracle (bypassing the Merant proxy). This works well.
Be sure to load test though - from memory, an earlier version of the driver had problems.
A newer driver is supplied with wM 6.X – sl53_cj33.jar. Haven’t had problem with that one. But of course, Sonam’s suggestion to test the driver first (on non-Production system!) is the prudent thing to do.
One more thing – we support a large number of customers, and definitely agree with wM’s suggestion that Cloudscape (in TN 4.X) not be used for production systems. They can’t be backed up on-line. They can’t shrink even if you deleted all transactions and conversations (CS limitation). The worst problem we had with them (although fairly uncommon) is some installations had the CS DB files just exploding (best way I can describe it. More than 1GB in 10 mins) for no apparent reason.
The SequeLink server is hard to set up, finicky sometimes. And it’s one more point-of-failure to worry about. Hence, our recommeded configuration is TN 4.X + Cloudscape for Test/Development server, and RDBMS + the sl53_cj3x.jar driver for Production server.
We are using external Oracle database, therefore Sequelink, also for development and QA environments.
However moving to production is causing lots of problems because IS is located in DMZ (protected LAN, behind firewalls). It happens that the firewall closes the TCP communication b/w IS and DB.
Moving from 5.2 to 5.3 with or without Sequelink server doesn’t change anything. Any idea how to manage/configure TN behind firewall?
Regards
Sorin,
Out of curiosity…
Why are you using Sequelink instead of the Oracle drivers?
Sorin,
That’s what I was thinking you meant. For webMethods 6.x, I don’t think there is any reason to not use the Oracle supplied thin JDBC driver with TN. It has always worked fine for me.
Over the years, I have only found 2 reasons to use Sequellink:
Your software vendor requires it. This has only happened to me twice in 10 years. The last vendor to do this to me I disqualified in a product evaluation.
Performance benefits by pre-fetching batches of rows on the server side while the client thinks it’s just fetching 1 row at a time. ( Modern versions of Oracle are good at providing pretty much the same benefit, so not much incentive here either )
In exchange, as you know, you have an extra license to pay for, an extra server-side configuration to maintain, and an extra layer of software to toubleshoot.
Try the thin JDBC driver.
Hi Mark,
unfortunately I’m still using 4.6 version (how could I forget to mention this!)
Reading out your answers in this forum I know you Mark have enough experience and I’d say very often found your replies very good & interesting. I’ll certainly try to see how TN is behaving with jdbc instead of sequelink drivers.
However I’m a little bit confuse: why wM is pushing sequelink if jdbc just does the same job? I really appreciate a comment over the query I posted (please notice the strange Join syntax in the from clause).
Thx and Regards
Sorin,
To your first question:
I don’t webMethods is pushing Sequelink for their current software. (6.x)
Vendors sometimes push Sequelink because it can make their job a little easier. Although you might think JDBC is a “standard”, like many specifications, it allows for a lot a variability from driver to driver. By forcing everyone to go through a “super driver” like SequelLink, they can avoid writing some additional code to account for the variation between drivers. You the customer pay the price in both dollars and effort when this happens.
But… these days, almost everyone expects good support for the vendor-supplied drivers from the big players: Microsoft SQL-server, Oracle and IBM DB2.
I know that for 6.x webMethods supports the Data Direct Connect drivers, (which come bundled with the 6.x installation) for the JDBC connection pools. These are not the same as Sequelink. I think Data Direct connect may be supported on 4.6 too, but I’d have to defer to someone else on that.
Can you try the Data Direct connect driver instead of Sequelink to see what happens?
Regards
ConnectJDBC is the defacto standard for current 6.x installations, and since it’s a type 4 driver, it doesn’t see nearly the installation problems that SequeLink had. Tn 4.6 was pretty picky about drivers, but even back in the day with 4.6, ConnectJDBC was offered to support clustered SQL Server installations, in which SequeLink could cause data loss. I’m not sure about the Oracle thin driver, and I don’t think that would be supported, but you can definitely get a driver that’s more user-friendly than SequeLink. Call webm support–it shouldn’t be a big deal to get setup with the updated driver and info.
“We can rebuild it. We have the technology.”
Tate
I’d agree with Tate, if you don’t already have the Connect JDBC driver, see if you can get it from webMethods support.
-Mark
I tried also to connect with sl53_cj32.jar driver (com.wm.dd.jdbc.oracle.OracleDriver). The results is the same or even worse than using Sequelink server…
Sorin,
Can you provide details on the error you get when you try to connect with the com.wm.dd.jdbc.oracle.OracleDriver?
-Mark
Now it is beginning to make sense!
If the network connection is lost, there isn’t much a JDBC driver or any other database client can do except attempt to restore the connection on a scheduled basis.
Closing a network connection through a firewall after a period of inactivity seems like a reasonable setup.
I would not usually expect to see a database connection going through a firewall. This is normally considered a security risk. Can we step back a little and discuss more generally what you are doing and why you are not going through the firewall with HTTP, FTP, etc?
I’ll be happy to discuss it.
I’m not an expert, but from my past experience, I think that for security reasons network people try to keep separately (as network) the webserver and the back office data (e.g. DB Server).
The webServer will have static IP/domain name and will be installed in a separate and very protected network, what they call DMZ, where firewalls will control HTTP/FTP/etc connection to and from Internet. I don’t know exactly how a DMZ is configured, but some how you will get also more protection to inside calls (that is versus/from back office systems).
The most of mine precedent installations of wM used to have IS inside local LAN, together with back-office, using proxy/NAT to access Internet.
In this case, however, IS is exposing several DSP pages as UserInterface, so for the network people is acting as a web server, so they decided to setup IS inside DMZ. It seams that is affecting DB connections especially through sequelink drivers (of any versions).
Oracle jdbc thin driver instead is acting a little bit better: while still having exceptions sometimes, it seams that connection pools using this driver will somehow open a new connection allowing IS to continue the work. Sequelink/DataDirect versions, once the connections are broken, failed to manage the situation and implied IS functionality becomes unavailable until IS completely restart…
I was wondering if there is a way to get control, programmatically, over the way wM manages the connection pool, so that I could intercept this exceptions and decide to clean the connection from the pool creating a new one instead! In your opinion, it’s this possible?
Please fill free to consider all my previous suppositions wrong, as there may be a way to correct/better configure the network and have everything working just fine!
Best Regards
Normally, the only IS you would find in a DMZ would be a reverse-proxy server.
Moving the “regular” IS inside the DMZ is not good.
Many companies do not even put webservers in the DMZ, allowing only proxy servers ( such as Microsofts ISA ) in the DMZ.
I would strongly recommend that you get the IS out of the DMZ and have the proxy environment configured to pass HTTP requests back and forth to it.
Regards
What you propose Mark makes a lot of sense to me (compliments!). Unfortunately network architecture/config does not depend on me, so all I can do about is try to explain and propose a inverse proxy solution.
However in this moment network people sustain this is a wrong behavior from wm app sever (i.e. IS!) and after months of trying to convince them otherwise I am finally facing “ the reality”: instead of accepting Process Monitor feature always unavailable I am trying to find a work around by catching exceptions and controlling pool connections. Hopefully this will be until upgrade to 6.1 version will consider a better network solution…
In few words my idea is to schedule an IS service that each 15min will try to execute sql for each connection in the pool, catching eventual exceptions. In base of exception code should decide to remove connection from the pool and eventually open a new one…
Any idea where to start?
I’m not sure what to suggest. Whatever you come up with is going to be a pretty unique setup.
Maybe others have some suggestions?