Aborted multi-machine transaction stays in BB

I have a distributed transaction going across 2 machines. I'm testing
out what happens when I abort the transaction. If I go into tmadmin,
and do a pt, I get the following printout. The problem is the
OK_ORATMSGRP in the first machine is specified being remote and as the
coord group and has a status of TMGACTIVE. The same group on the other
machine (where the group is local) has a status of TMGABORTED. Now this
is exactly the same group, but one machine thinks it is TMGACTIVE and
the other machine thinks it's TMGABORTED (this is the correct status).
The transaction will remain in the bulletin board until the timeout
value expires. This shouldn't happen. I thought as soon as I called
tx_rollback the whole transaction should be aborted. What is going on?
This is causing a problem because before we didn't specify a transaction
timeout and the BB was being filled with Aborted transactions that had
that one group marked as TMGACTIVE when it shouldn't have been.
Subsequently the BB became full and new transactions couldn't be
started.
index=2 gtrid=x0 x3d25495d xa61
: Machine id: IRGVELOCITY, Transaction status: TMGABORTED
Group count: 2, timeout: 30, transaction timed out
Known participants:
group: OK_ORATMSGRP, status: TMGACTIVE, remote, coord
group: OK_FXNTGRP, status: TMGABORTED, local
Machine id: CDDEV, Transaction status: TMGABORTED
Group count: 4, timeout: 30, transaction timed out
Known participants:
group: OK_ORATMSGRP, status: TMGABORTED, local, coord
group: OK_MQTMSGRP, status: TMGABORTED, local
group: OK_TUXTMSGRP, status: TMGABORTED, local
group: OK_FXNTGRP, status: TMGABORTED, remote

So it would be fair to say that as a rule, you should always specify a reasonably
low transaction timeout so the BB doesn't become full of aborted transactions? I
still think that an aborted transaction should be removed from the BB at some point
(especially for aborted transactions with no timeout value). Maybe there should
be a different timeout value, separate from the transaction timeout value that determines
how long an aborted transaction should remain in the bulletin board.
Peter Holditch <[email protected]> wrote:
Frank,
The timeout is in there to prevent the scenario where a transaction aborts
and
is deleted, and then another message on behalf of the same transaction arrives.
With the records having been deleted, the system has no option but to
create a
new GTT entry for the apparently new transaction and then wait for it to
time out.
I know that this timeout is also in the WLS implementation. I do not believe
that it is infinite!
Note also that the only trandactions that will ever hit the TLOG are ones
that
have decided to commit, by which time this timeout is not relevant any more.
That's my recollection of the reasoning, I hope it sheds some light...
Regards,
Peter.
Frank Clarijs wrote:
Failed (rolled back) distributed transactions remain in the BB until
timeout, by design. I'm not sure why, but this behaviour is the one BEA
(or its predecessors) explicitely implemented.
This has of course strange consequences: if your application aborts
transactions quite often, either you need a big TLOG and even higher
setting of "maximum global transactions" in the UBB (or you can't
restart your application after a crash), or you need to set the timeout
quite low, with all risks from that...
Don't worry, it's even worse in Weblogic, where the bean for an aborted
transaction never disappears, by design.
Frank
Anthony Fryer wrote:
I have a distributed transaction going across 2 machines. I'm testing
out what happens when I abort the transaction. If I go into tmadmin,
and do a pt, I get the following printout. The problem is the
OK_ORATMSGRP in the first machine is specified being remote and as the
coord group and has a status of TMGACTIVE. The same group on the other
machine (where the group is local) has a status of TMGABORTED. Now
this is exactly the same group, but one machine thinks it is TMGACTIVE
and the other machine thinks it's TMGABORTED (this is the correct
status). The transaction will remain in the bulletin board until the
timeout value expires. This shouldn't happen. I thought as soon as I
called tx_rollback the whole transaction should be aborted. What is
going on? This is causing a problem because before we didn't specifya
transaction timeout and the BB was being filled with Aborted
transactions that had that one group marked as TMGACTIVE when it
shouldn't have been. Subsequently the BB became full and new
transactions couldn't be started.
index=2 gtrid=x0 x3d25495d xa61
: Machine id: IRGVELOCITY, Transaction status: TMGABORTED
Group count: 2, timeout: 30, transaction timed out
Known participants:
group: OK_ORATMSGRP, status: TMGACTIVE, remote, coord
group: OK_FXNTGRP, status: TMGABORTED, local
Machine id: CDDEV, Transaction status: TMGABORTED
Group count: 4, timeout: 30, transaction timed out
Known participants:
group: OK_ORATMSGRP, status: TMGABORTED, local, coord
group: OK_MQTMSGRP, status: TMGABORTED, local
group: OK_TUXTMSGRP, status: TMGABORTED, local
group: OK_FXNTGRP, status: TMGABORTED, remote

Similar Messages

Adobe After Effects CC 2014 Multi-machine rendering does not create RCF after Collect-File

This was working in the previous release. After going through all of the steps to perform a multi-machine render, not only does it say it is limiting it to 5 machines instead of the usual 99 (and I have over 100 cores on the render farm) but "Collect Files" (File/Dependencies/Collect Files) does not create the Render Control File (RCF) that kicks off the process on the other machines, like it did before and is supposed to do.
This release was supposed to fix the 32 core limitation on render nodes (my server has 48), but now does nothing at all.
Anyway, I assume there is a check box somewhere that says "do something" that hasn't been checked and another that says "Maybe my machine has more than 5 cores", and I would appreciate being steered in the direction of these new mis-features.
Just as background, the 48 core server does show the little TV screen that monitors the watch folder, and I know for a fact that process looks for the "RCF" file. Looking at the directory specified for "collect files", I can see that no RCF file is created, only 1 file, a report.txt file which contains this information is produced:
Report created:
7/9/2014 1:54:12 AM
Project name: highup.aep
Source files collected to:
P:\temp\highup folder
Source files collected: All
Collected comps:
Thomas Falke - High Again High On Emotion)
Number of collected files: 1
Size of collected files: 63.4 MB
Collected source files:
C:\Users\Jeff K\Videos\Down\2014-2015\Thomas Falke - High Again High On Emotion).mp4
Rendering plug-ins:
Classic 3D
Effects used: 1
Effect: Detail-preserving Upscale

The dialog box has been slightly refactored. You must click "Allow watch folder rendering" and then all is good. Hope this helps anyone else. Thanks to Rafil at customer support chat for this.

SET CHAINED command not allowed within multi-statement transaction

Hi,
i need to do one transaction and i am setting autocommit(false) and
once i am don ewith my 2 inserts i am settins it to true.
i have like 20 thousand rows and when it is working fine for some inserts but after that it is throwing
"SET CHAINED command not allowed within multi-statement transaction".
can anyone please help me.

You can find some information here:
http://publib.boulder.ibm.com/infocenter/wsphelp/index.jsp?topic=/com.ibm.websphere.nd.doc/info/ae/ae/rtrb_dsaccess.html
it's first problem for sybase DB

After Effects multi machine render frames have different settings.

I am using AE multi machine render settings over a network drive with 2 machines. Comp includes elements from Video Copilot plugins. The ambient occlusion settings appear to differ despite the same AE project being open on both machines. Certain elements in the comp are rending with slightly different settings for each machine, creating a flicker effect when played back at the intended video fps. Both machines have all the necessary elements installed locally to complete an exact identical render. What am I doing wrong?

Well, have you actually tested it? E3D will e.g. truncate the number of samples to the highest possible value internally, but still allow you to overcrank sliders. This sounds like exactly that - you may have set 32 samples or whatever, but internally only 16 or 8 are used. You can only find that out by actually interactively opening the project and changing those values.
Mylenium

How to Abort all opened Transactions at once?

Hi,
Is there any way to abort all open transactions at once ?
Thanks for any help.
Thanks,
reddy

Reddy,
No, you need to call Transaction.abort() on each transaction individually.
Regards.
Linda

AE CC Multi-Machine Rendering: Still asking for Adobe ID on Render Engine App Launch

Hi everyone,
I'm sure this is a fairly simple fix, but I have looked for quite a while through blogs and AE documentation and I just can't get it figured out.
I'm trying to set up my first multi-machine render farm with AE CC. All machines for the render farm are running Windows 7 and are networked via Gigabit Ethernet Switch. I understand that I can have CC installed on 2 computers with my membership and then I can install a number of Render-Only instances of AE. The problem is I am trying to set up one of my 3 machines in render-only, yet I am still being asked to sign in with my Adobe ID when I launch the Render Engine App. I can't sign in a third time or it will ask me if I want to sign out of all of the other computers that are licensed under my name. I've attached screenshots of what it shows.
I have already put the ae_render_only_node.txt file in the correct place as per the Adobe documentation (C:\Users\<myusername>\My Documents)
Any thoughts? I'm guessing I'm doing something dumb, but I just can't pinpoint it.
Thanks for any help!

One thing to check: in Windows Explorer, go to the Tools menu > Folder Options. Disable "Hide extensions for known file types". Having done so, is the file still named "ae_render_only_node.txt", or does it have an extra .txt on the end?
If not, what happens if you move the file to the alternate location? C:\Users\Public\Public Documents\Adobe

My Time Machine Backup Stays Forever

My Time Machine Backup Stays Forever, it says backingup and no progress, some times indexing stays a whole day ??!!

Were you able to back up before, or is this a new setup?
What are you backing-up to (external HD, Time Capsule, or what)?
It sounds like the backups may be damaged; try repairing them, per #A5 in Time Machine - Troubleshooting.

Multi-Machine Blueprint- How to share IPs, passwords, etc..

I am trying to configure a multi-machine deployment in Application Director. I want the machines to have each others IPs, db passwords, and things like that. So they can all run together in a little environment hosting apps and websites. I am looking for help on how to share these variables with all the VMs in the blueprint. I see some out of the box examples in App D, but I don't find a properties dictionary or any docs that explain how to accomplish this. I have attached a pic to show what VMware has done out of the box. Notice the "appserver_node_ips" and the text in the value column.
Does anybody know how to set this up or where to get info that explains these properties and their options?

You can use component dependencies to share property values between components. E.g.
Say I have a node which will have 2 app components on it.
Component c1 has properties
p1
p2
Component c2 has properties
p3
p4
p1copy
p2copy
If you look in the upper right of the Blueprint builder there is a chain icon which lets you define dependencies between components (either on the same nodes or different ones)
With a dependency in place between my example components (c2 depends on c1) we can then bind the p1copy property value in c2 to the value in c1 in the following notation
self:c1.p1
This expression breaks down as
<node name>:<component name>:<property name>
Quite simple and very powerful
Also, there are some built-ins for each node such as
self:ip
self:hostname
etc.
This is all mostly in the docs but you'll wrap your head around it pretty quickly if you play about with it

How to request multi-machine blueprint information through rest api.

We are building a portal in front of vRealize (don't ask me why) and I need to dynamically create multi-machine json files to request catalog items.
If I connect through the api I can request the Blueprint details but not the additional blueprints it makes use of.
Does anyone know how to request these additional information out of the rest api catalog?
Thx in advance.

The following method, unlike those above, will work. It requires you to know the MIME type of the binary file you wish to upload. This example uploads a PDF document.
uploader.cfm
<cfset restInitApplication(expandPath("."))>
<cfhttp url="http://localhost:8500/rest/myRestApp/myService" method="post">
<cfhttpparam file="C:\temp\testDoc.pdf" mimetype="application/pdf" name="myFile" type="file">
</cfhttp>
service.cfc
<cfcomponent rest="true" restpath="/myService">
<cffunction name="getFile" access="remote" httpMethod="POST" returntype="string">
<cffile action="upload" accept="application/pdf" filefield="myFile" nameconflict="overwrite" destination="C:\Users\BKBK\Desktop\testFile.pdf">
<cfreturn "File uploaded!">
</cffunction>
</cfcomponent>

Multi-machine configuration

how do i configure the Master and other machines in a Multimachine configuration.
i have 10 machines.
In RESOURCES
MASTER appsvr_1 appsvr_2
MACHINES
byspa01 LMID=appsvr_1
byspa02 LMID=appsvr_2
byspa02 LMID=appsvr_3
Is this a right configuration?
Or should LMID be same as Master? PLease advise.
Thanks
sushil.

This is a very general question. I'd recommend the documentation section at
http://e-docs.bea.com/tuxedo/tux81/ads/adfig.htm#1253022 for a good overview
of how to set up a multi-machine configuration file.
Also, you may want to work with some of the Tuxedo sample applications to
get a feel for how to set up a Tuxedo configuration file. The bankapp
sample application located in $TUXDIR/samples/atmi/bankapp has a 2-machine
configuration and is a good application to gain familiarity with Tuxedo in a
multiple machine environment.
<Sushil More> wrote in message news:[email protected]..
how do i configure the Master and other machines in a Multimachineconfiguration.
>
i have 10 machines.
In RESOURCES
MASTER appsvr_1 appsvr_2
MACHINES
byspa01 LMID=appsvr_1
byspa02 LMID=appsvr_2
byspa02 LMID=appsvr_3
Is this a right configuration?
Or should LMID be same as Master? PLease advise.
Thanks
sushil.

Multi-machine sequence missing from AE?

Hi there
after i installed the script BG Render, not sure what happened but multi-machine sequence becomes missing from AE as it says on startup that there is an output module missing from after effects, really need to solve this as I am running out of time, if anyone knows how to repair this or a solution please help!
Thanks

If you have deleted one of the default Output Module or Render Settings templates, there are a couple of ways to get them back.
On a different computer that has the templates, save the templates from the template editor dialogs (Edit > Templates), copy the resulting .aom or .ars file to the machine where the template is missing, and load the saved templates.
Trash the preferences file. If you're using After Effects CC, you only need to delete the Adobe After Effects 12.0 Prefs-indep-output.txt or Adobe After Effects 12.0 Prefs-indep-render.txt file. For earlier versions of After Effects, you'll need to trash the whole main prefs file (for CS6, Adobe After Effects 11.0-x64 Prefs.txt). To find these files, see the Reveal Preferences section in After Effects Help.

Why does multi-machine rendering work so poorly?

more specifically, watch folder rendering with multi-machines is pointless because skipping a frame seeming to take as long as rendering the frame, so one machine can never catch up to the other. why does skipping a frame take some long? i mean it's been in the software for like 10 years. is this not considered important?

These two features are not intended to be used together.
Render Multiple Frames Simultaneously multiprocessing is intended to turn a single computer into a self-contained render farm of sorts. Multi-machine rendering is intended to use multiple computers in a render farm. Attempting to use these two items together is mixing two different organizational schemes for managing the rendering of frames.

Sybase: SET CHAINED command not allowed within multi-statement transaction

Hello,
I'm getting the error message "SET CHAINED command not allowed within
multi-statement transaction" for CMP Entity beans against Sybase.
The errors appear in my jdbc.log in this order:
010SM: This database does not support the initial proposed set of
capabilities, retrying.) SQLState(010SM)
JZ0EM: End of data.
JZ0SJ: Metadata accessor information was not found on this database.
Please install the required tables
as mentioned in the jConnect documentation.
010SL: Out-of-date metadata accessor information was found on this
database. Ask your database administrat
or to load the latest scripts.) SQLState(010SL)
SQLState(ZZZZZ) vendor code(226)
com.sybase.jdbc2.jdbc.SybSQLException: SET CHAINED command not allowed
within multi-statement transaction.
I'm using JConnect 5.5, WebLogic 6.1 sp3, Solaris 8, and Sybase 11.
The weird thing is, the app works on a box running 6.1 sp2 and Win2k
Prof.
Any ideas?
Thanks!
Dan

Hi Dan,
There is a patch for this sybase problem, please contact [email protected] to
get a temp patch.
sree
"Dan Blaner" <[email protected]> wrote in message
news:[email protected]..
Hello,
I'm getting the error message "SET CHAINED command not allowed within
multi-statement transaction" for CMP Entity beans against Sybase.
The errors appear in my jdbc.log in this order:
010SM: This database does not support the initial proposed set of
capabilities, retrying.) SQLState(010SM)
JZ0EM: End of data.
JZ0SJ: Metadata accessor information was not found on this database.
Please install the required tables
as mentioned in the jConnect documentation.
010SL: Out-of-date metadata accessor information was found on this
database. Ask your database administrat
or to load the latest scripts.) SQLState(010SL)
SQLState(ZZZZZ) vendor code(226)
com.sybase.jdbc2.jdbc.SybSQLException: SET CHAINED command not allowed
within multi-statement transaction.
I'm using JConnect 5.5, WebLogic 6.1 sp3, Solaris 8, and Sybase 11.
The weird thing is, the app works on a box running 6.1 sp2 and Win2k
Prof.
Any ideas?
Thanks!
Dan

Multi-"Machine" Activation Question: Virtual Machines

Hi,
I am developing plugins and have the need to test with a wide range of Photoshop versions on a wide range of operating systems.
At the moment I'm accomplishing this through the use of VMware virtual machines. I have a powerful 64 bit host workstation running Vista x64 and have created virtual machines for XP Pro, Vista 32 bit, Win7 RC, and Win7 retail. Virtualization saves me from having to build a lab full of test hardware.
I have the desire to test with versions 6 through CS4... I haven't yet gone through the process of installing all the versions of Photoshop on all these VMs; just a few, which is probably covered under the 2 activations limit... My question is this:
Will Adobe consider it a separate activation for each virtual machine? Can they be "sweet talked" into allowing more? I am running them all on one computer, and (generally speaking) using only one at a time.
From what I can see the software will not realize this is one and the same computer, since the virtualization environment does a good job of convincing the OS that it's a completely separate computer... I suppose I will forge ahead and call Adobe... Adobe phone contact is not touted to be much of an easy or rewarding thing from what I hear.
Any advice on how to proceed with setting up a decent testing environment to support long-term multi-platform testing would be great.
Thanks in advance.
-Noel

I would read this attached page. We are pretty much users like you and cannot speak for adobe or how they might react to your question. Calling them is the only option.
http://www.adobe.com/aboutadobe/volumelicensing/

Aborted time machine backup

I started a time machine backup. I aborted within 5 seconds by unplugging the external HD. I restarted and the it won't let me startup normally. I forgot that I have pictures that need to be copied first. Surely the HD can't be wipedclean in a few seconds. Is there a way to restart my macbook pro and get these pics? Thanks

Mac OS X alone has over 380,000 files in it and takes up over 10GB. If your drive is showing as "all but empty", the restore must have erased the drive, then began to restore files.
I expect that means those special files you were looking for are gone. unless they were already on the Backup.
Surely the HD can't be wipedclean in a few seconds
No, it can't be. But the directory can be removed, effectively emptying the drive. The data blocks may still mostly be there, if you could rescue them to another drive. But with the Directory gone, you could recover 380,000 files and still have no names for anything.

Aborted multi-machine transaction stays in BB

Similar Messages

Maybe you are looking for