Cloverleaf 5.6

[John Hamilton]

Any one out there using this?

I’m thinking of upgrading and wanted to get input from as many sources as I can too see about any issues people have run into.

It has several features I would really like to get here.

Topic

[James Cobane]

John,

We are just in the beginning stages of preparing to upgrade to 5.6 (currently running 5.4.1 on AIX 5.3 ML4).  When we start testing, I’ll keep you appraised of what we find.

With that said, in reviewing some of the new features and enhancements in 5.6, it looks like there are some nice improvements.  I especially like the changes to the Xlate GUI.

Have fun.

Jim Cobane

Henry Ford Health

[Bill Marsch]

We may have an issue with testing as we have the version 5.6 on the same server with 5.3.  it is on an AIX 5.2 UNIX server where our current test sites and prod sites are under the same root or /quovadx/qdx5.3/integrator.  We also have our crontabs set with commands to cd $HCISITEDIR.  This caused an issue while doing the install as when root was set to the new version /quovadx/qdx5.6/integrator our current production crontabs failed as when it did the cd’s it did not find the proper directory to execute scripts.  How can you be running both versions testing 5.6 and running 5.3 production processing?  Anyone have any suggestions on this?  Everything is back to normal so we can do production but how does one version survive with the other under our conditions?

[James Cobane]

Bill,

One of the nice things about Cloverleaf is that you can run multiple versions simultaneously (as long as those versions are supported under the same OS, etc).  As long as you setroot to the desired version, you should be able to function and test.  You may need to modify the hci user’s .profile to reflect the older (production) root so that your environment variables are set appropriately for the existing cron scripts, etc., but both versions should co-exist.  If you’re having specific issues, I would recommend talking to support, and they should be able to help you out.

Jim Cobane

Henry Ford Health

[Mark Thompson]

Bill,

The AIX install scripts modify the /etc/environment file.  Immediately after install, we modify the FPATH and QUOVADX_INSTALL_DIR entries back to their previous value.  That way the existing cron scripts continue to run.  You can setroot to /yourpath/qdx5.6/integrator from a command line to do your testing.

Cloverleaf 5.6 introduces a new wrinkle.  The QUOVADX_INSTALL_DIR environment variable has been replaced by CL_INSTALL_DIR.  Not sure if that is political or technical, but if you are relying on the environment variable it is something to check.

- Mark Thompson
HealthPartners

[Bob Richardson]

Greetings,

Folks, we have run evaluation tests using Cloverleaf 5.6 on AIX5.2 and found performance degradation from 13% to 18% from our current Cloverleaf 5.3 Rev3 engine.  We plan to run an evaluation on AIX5.3 hoping for performance that at least equals (preferably exceeds) our current version on AIX5.2.  We use BULKCOPY heavily in our xlates and have numerous pre/post TCL stacks in multiple connections spread throughout mulitiple sites – there is a site re-balancing task on our plate that is tabled for the moment.

Please let us know if you experience performance improvement and/or degradation on your platform with an indication of the complexity of your implementation.

For some good stuff:  the IDE loads much faster and is more responsive!

A definite plus!

Enjoy.

[Bill Marsch]

Response time and performance is a real big issue with us.  We are in the process of upgrading to 5.6 and we have noticed an increase in the time for the interfaces to start as well as process.  We have a fair share of preprocessing TCL’s and translates with bulkcopy though most of the xlates have path copies combined with specific copy commands.  We would not like having to go backwards.  Is there anything that can be done to increase performance.

[Bob Richardson]

Greetings,

We have discovered through some evaluation testing that the new Cloverleaf 5.6 Integrator performs a thread count for each site and compares that against your new license for the integrator.  Quovadx is now enforcing a thread limit (connection boxes if you will).  This audit is done at startup time.  This was confirmed with Quovadx R & D at the time of our evaluation trials.

Enjoy.

[Chris Brossette]

What are the server configs?  Processors & memory?  Could the 5.6 upgrade require more memory to execute?  We are on IBM JS-20 Blade servers and would like to keep these for another couple of years before upgrading.  We have not had any issues with 5.4 and our next upgrade will be to 5.6.

Thanks…..

[Bill Marsch]

We have an unlimited license which does not have the thread count.  Maybe it still does it though I would think taht some switch would tell it not to when the license is unlimited.

[Bob Richardson]

Bill and company,

We have an unlimited license as well but the audit executes.

I agree that the logic could bypass this audit for customers whose license is unlimited.  Maybe a Rev 1?

BobR

[Rob Abbott]

All – just to let you know, Bob submitted an enhancement request for the audit bypass.  This is “on the list” and although I can’t promise anything I hope to see it addressed in a 5.6 rev.

Rob Abbott
Cloverleaf Emeritus

[Todd Lundstedt]

Rob, et al,

Can you indicate whether the audit is performed at process startup,  thread start, or what?

Rob, thanks for the enchancement request update.  Let us know when it is officially on a Rev fix list, please.  Depending on the audit frequency and overhead, this will weigh heavily on decisions being made at our facility.

[Max Drown (Infor)]

I hope it’s OK if I use this thread for general 5.6 questions, and my apologies  if this has been discussed elsewhere already (if so, please point me).

In 5.6, where are the NetConfig notes being stored? Is there a way to query these notes from say a webpage or tcl/perl script?

-- Max Drown (Infor)

[Bill Marsch]

Though there were a few issues along the way we are pleased with the results of the initial testing and omplementation of version 5.6 of cloverleaf.  It was tricky with having not all sites go at the same time and having cron jobs run for sites under version 5.3 and version 5.6 for live sites where we had to set the appropriate root for them to execute.  We especially like the changes to SMAT as to the changes that were made for veiwing message files on the fly and the veiwing of the ebcdic in smat.  We also like the new editor for Xlates and the handle an look of the Xlate edit screens.  We did find the issue with starting the threads due to the thread counting that was noted before in this posting.  We are looking forward to Charlie’s Alert changes in a future release.

[Mark Thompson]

Max,

The notes for each thread are stored in

$HCIROOT/$SITE/notes/NetConfig/$PROCESS/$THREADNAME.notes

- Mark Thompson
HealthPartners

[Todd Lundstedt]

Bill (or Charlie),

Can you give more detail on the “Alert changes” mentioned a couple posts up?

edit: ah.. never mind.. I remember the wish list post, now.

[Rob Abbott]

Here’s what I got on thread count audit.  The thread count is executed on the following conditions:

[Todd Lundstedt]

Thanks Rob, that helps..

So a site that has 12 processes will run the audit 12 times during a failover (not good, IMHO) covering all autostart threads, and then again for each thread start or process cycle after that.

Please clarify the count routine, I assume each audit is counting ALL defined protocol threads in ALL sites.

What kind of overhead do you expect this to cause on large sites or large multi-site systems: would we even notice the overhead with 12 processes in one site?

During a failover, would the audits be single threaded, or do they run in parallel (given enough CPU separation).

My upper management is thrilled with our failover time, now… it only takes 3 minutes to failover, and another minute to catch up during peak periods.  Our previous implementation was 15-25 minutes for the failover, and 45-75 minutes to catch up all interfaces.  

Thanks again for your input.

[Rob Abbott]

The thread count routine is counting all *running* threads for all sites in the root.  My understanding is it does this:

– find all sites in the current root

– for each site found, interrogate shared memory for running thread total

I don’t have any metrics on overhead.  I believe (unconfirmed) that the audit process is single-threaded.

Rob Abbott
Cloverleaf Emeritus

[Max Drown (Infor)]

Mark Thompson wrote:

Max,

The notes for each thread are stored in

$HCIROOT/$SITE/notes/NetConfig/$PROCESS/$THREADNAME.notes

What are the limitations on the Notes field?

-- Max Drown (Infor)

[Todd Lundstedt]

Is it a good assumption that the process startup is dependant on the successful completion of the audit process?  If so, and if your belief is true, then process startup just became single threaded (to a point).

Please keep us informed of the intended release/version and estimated GA date for improvements to this process.

Thanks!

[Charlie Bursell]

Max:

Notes are under the directory named notes  😀

Look in the site directory

[Bill Marsch]

With this new version the two greatest items for testing and research has been the enhancements to SMAT.  The vewing of the current save to inbound outbound files and not having to do a cycle save of them before viewing them or having to shut down the interface.  The other item is the enhancement made to HL7 veing in SMAT where you can select the version, varient, display field names and the detail level of the viewing of the message.  This is cool!

[Todd Lundstedt]

That last one you mentioned, formatted display of records in SMAT, is actually available in 5.5, along with searching inbound, outbound and actions in a transalte (nice!).  If you are a large site and want to avoid the whole “audit thread count” hassle until they get it fixed for unlimitted thread counts, upgrading to 5.5 will get you most of the new cool stuff.  

Though, the live SMAT files thing is a very nice thing….

[Rob Abbott]

Hi all, wanted to give a status on this.

R&D has reviewed the license checking at startup.  It turns out that we have been chasing the wrong issue here.  

If you have an unlimited license in 5.6, the thread audit count is not performed.

There’s something else going on during startup.  We are diligently investigating other causes for a performance hit.  I’ll try to keep this thread updated with results.

Rob Abbott
Cloverleaf Emeritus

[Max Drown (Infor)]

In previous versions we were using recovery_33 scripts (sendOK_save, resend_ob_data, check_ack). Do we no longer use any or all of those scripts? How should the NetConfig look for a “standard” outbound tcp/ip connection?

-- Max Drown (Infor)

[John Hamilton]

As I understand it that is your choice.

You can use them or there is a new built in function that preforms the same task.

With that said I have seen several posting where people are having problems with the state 14 messages not being cleaned up correctly I don’t know all the detail on those but be aware there may be an issue with one way or the other on high volume threads.

[Max Drown (Infor)]

Work around for the duplicate state 14 messages: http://clovertech.infor.com/viewtopic.php?t=2640

-- Max Drown (Infor)

[Bill Bertera]

Any word on when we could expect a Rev for 5.6?

[Rob Abbott]

5.6 rev1 is planned for late April.  I do not have a hard date.  This timeframe is subject to change.

The following fixes are PLANNED.  Do not take this as written in stone.  I share this information in good faith as an attempt to provide good communication to everyone.  

[Jim Kosloskey]

Rob,

Thank you very much for posting this information.

It has come in handy as we are in the planning stages and we have been talking about how we will handle the situation.

Jim Kosloskey

email: jim.kosloskey@jim-kosloskey.com 29+ years Cloverleaf, 59 years IT - old fart.

[Russ Ross]

Rob:

As Jim Kosloskey pointed out we are fitting in the upgrade to QDX 5.6 (AIX 5.3) from QDX 5.2 (AXI 5.2) between what cracks in time we can pry loose.

That means rev1 might be out by the time we get there but if we get some sites migrated to QDX 5.6 before rev1 comes out I was wondering about one of your rev1 fixes.

You mentioned hcidbconvert does not work on AIX along with comments about the Ramia DB.

My question is when migrating from QDX 5.2 (AIX 5.2) to QDX 5.6 (AIX 5.3) will the Rami DB be upgrading causing hcirootcopy to automatically run a broken hcidbconvert.

If not do I have to manually run a broken hcidbconvert and if it is broken what is the work around prior to rev1?

Russ Ross
RussRoss318@gmail.com

[Rob Abbott]

I sort of think this falls into the category of asking for details *cough*

I really don’t want to comment further as I don’t want to set expectations inappropriately.

In any case, here is what I see regarding this issue.  The AIX raima upgrade in rev1 fixes this.

Quote:

With Raima downgraded to 4.5 on AIX, we found hcidbconvert does not work with the following error:

$ hcidbconvert Raima Database Manager 4.5 [Build 1] RDM UNIX Database Export Utility Copyright (C) 1984-1996, Raima Corporation, All Rights Reserved Database Import Utility – RDM Server 8.0.807C [22-Oct-2007] A Raima Database Manager Utility Copyright (c) 1992-2007 Birdstep Technology. All Rights Reserved. SYSTEM/OS error: -905 error opening data or key file C errno = 2: No such file or directory press to continue Recovery data export failed.

The work-around: cd to the $HCISITEDIR/exec/databases directory, then run hcidbconvert. Make sure db server is running first using hcisitectl -s d

Rob Abbott
Cloverleaf Emeritus

[Bill Marsch]

it appears that with version 5.6 when selecting lookup table configurator the .tbl extension is being enforced and will list only *.tbl entries unlike 5.3. As well when in the translate configurator when selecting tables it will only list *.tbl entries and in the testing tool if a translate or tcl script uses a table that does not have the *.tbl extension then it will abend with “error “site table $tableName does not exist””  Please note since this was excluded from the release notes and for some of us that did not adhere to the .tbl extension for our tables in previous releases you must rename your tables and place a “.tbl” extension on them.

[Rob Abbott]

Thanks Bill, I wasn’t aware of this, I will submit a doc-bug so that the release notes get updated.

Rob Abbott
Cloverleaf Emeritus

[Bob Richardson]

Greetings,

Folks, we have CIS5.6 loaded on our AIX 5.2 Test Server.  In looking at the documentation for hcidbconvert and plowing down to details on hcisitectl -d it appears that this is an issue on a Windows platform only and not something for an AIX shop to worry about.

Is this true?

Please confirm.

Thanks in advance.

[Bill Marsch]

The extention on the table files what will happen if you have a mix bag where the file name has no *.tbl extension and the entry in the translate does have the *.tbl and the other condition is where the file has a *.tbl extention and the entry in the translate does not?

[Kevan Riley]

>I hope it’s OK if I use this thread for general 5.6 questions, and my

>apologies  if this has been discussed elsewhere already (if so, please

>point me).

>In 5.6, where are the NetConfig notes being stored? Is there a way to

>query these notes from say a webpage or tcl/perl script?

The notes are in a directory off the site called “notes”.  Under “notes” is a NetConfig directory containing subdirectoies by process name.  Then each thread that has notes, have their own file.  For instance, for conn_1 there is a file called “thread_conn_1.notes”.  This is a plain text file.  This whole directory structure does not appear to exist until a note has been created.

I hope this helps

Kevan

[Edit:  Sorry for re-posting this information.  I did not notice until now that this had been answer twice already in this thread.]

[Bill Marsch]

are there any command line commands for debuging your java rmi in unix?

[Rob Abbott]

Rob Abbott wrote:

Hi all, wanted to give a status on this.

R&D has reviewed the license checking at startup.  It turns out that we have been chasing the wrong issue here.  

If you have an unlimited license in 5.6, the thread audit count is not performed.

There’s something else going on during startup.  We are diligently investigating other causes for a performance hit.  I’ll try to keep this thread updated with results.

After working with customers that noted performance hits, it seems that there are no discernible performance issues in 5.6.  If anything, the engine is performing slightly faster in 5.6 when compared to 5.4.1 or 5.5.

I think the startup slowdown that was observed was due to a DNS validation that occurs at thread startup.  

Starting in 5.6 rev1,  hostnames in NetConfig (e.g. as part of a TCP/IP client protocol configuration) will no longer be validated in the engine.  Hostnames will be validated when NetConfig is saved or when hcinetcheck is executed.

Hope this clears up any performance concerns.

Rob Abbott
Cloverleaf Emeritus

[Author]

Replies