Using 2 different Switches for illustrator

[PdFUser5000]

Using only one Illustrator is sometimes not enough. Since i have 2 Switches, is it possible to use the other one for Illustrator also?

Is there a way to send files to the other illustrator only when the first one is busy? Or should i just start adding a counter to the metadata and sending odd numbers to one switch and even numbers to the other? Are there any other options for this?

[freddyp]

When you add 2 outgoing connections to "Hold job" Switch will send 1 job here, 1 job there, 1 job here, etc. In this way the number of jobs will be equally split over the outgoing connections.

That is not always the best choice because not all jobs take equally long. To make it a bit more intelligent, have a look at the properties of the outgoing connection of "Hold job". Use "Folder constraint" and set the number of jobs to 1. The load balancing will now happen based on which outgoing connection becomes free first.

[PdFUser5000]

When you add 2 outgoing connections to "Hold job" Switch will send 1 job here, 1 job there, 1 job here, etc. In this way the number of jobs will be equally split over the outgoing connections.

That is not always the best choice because not all jobs take equally long. To make it a bit more intelligent, have a look at the properties of the outgoing connection of "Hold job". Use "Folder constraint" and set the number of jobs to 1. The load balancing will now happen based on which outgoing connection becomes free first.

Thanks, that worked! One question: What happens to files when the second Switch instance is not working for example? are they sent to problem jobs? Is there a way to create a safety net, which in the case of only one switch working does not send jobs to the send to switch portal?

[freddyp]

It depends how you transfer the jobs to the other Switch and what it is that is not running.

If you drop the job in a local folder that is a remote folder for the other Switch, the job will be stuck in that folder. As it is stuck "Hold job" will route all the other jobs to the other local instance.

If you drop the job in a remote folder and the remote computer is down, you will get errors, and if the remote computer is up but Switch is down then you will have the same situation as in the first case.

My advice: use the Switch2Switch bundle of apps. This is Switch-to-Switch communication, there are no folders involved. In other words, when the remote Switch cannot be reached for whatever reason, the job will follow the error connection and you can continue to process it locally.

[PdFUser5000]

It depends how you transfer the jobs to the other Switch and what it is that is not running.

If you drop the job in a local folder that is a remote folder for the other Switch, the job will be stuck in that folder. As it is stuck "Hold job" will route all the other jobs to the other local instance.

If you drop the job in a remote folder and the remote computer is down, you will get errors, and if the remote computer is up but Switch is down then you will have the same situation as in the first case.

My advice: use the Switch2Switch bundle of apps. This is Switch-to-Switch communication, there are no folders involved. In other words, when the remote Switch cannot be reached for whatever reason, the job will follow the error connection and you can continue to process it locally.

Switch2Switch works perfectly, thanks again!

[PdFUser5000]

When you add 2 outgoing connections to "Hold job" Switch will send 1 job here, 1 job there, 1 job here, etc. In this way the number of jobs will be equally split over the outgoing connections.

That is not always the best choice because not all jobs take equally long. To make it a bit more intelligent, have a look at the properties of the outgoing connection of "Hold job". Use "Folder constraint" and set the number of jobs to 1. The load balancing will now happen based on which outgoing connection becomes free first.

Something still isn't quite right. I set folder constraint to 1 on both outgoing connections of the hold job element. I also set the target folder to the same folder( Timestamp Input) on both connections. But it seems switch still moves jobs cyclically between the 2 connections eg when the first Switches illustrator input folder is empty, Switch still sends it to the second Switch. Am i missing something here, or is it supposed to be like that?

Capture.PNG (28.98 KiB) Viewed 54843 times

[freddyp]

There is no need to set a target folder and the second Hold job element is not necessary. This is the principle:

Screen Shot 2021-05-10 at 11.39.57.png (27.25 KiB) Viewed 54839 times

[PdFUser5000]

There is no need to set a target folder and the second Hold job element is not necessary. This is the principle:
Screen Shot 2021-05-10 at 11.39.57.png

Made the fixes. Switch still keeps sending jobs cyclically between the two connections.

[freddyp]

Are you sure you set the folder constraints on the outgoing connections?

Put the connections after Local and Remote on hold and place 5 input jobs in the input folder. The folders should each get 1 job and there should be 3 jobs waiting before "Hold job". Release one of the two connection and see what happens. The released folder should now get a job after the previous one was processed and the number of waiting jobs drops to 2. If that is not the case, then there is something wrong with the settings. I have used this method often enough to know that it works.

Put differently, the number of jobs in Local and Remote should never be higher than 1 (unless the sending to the remote Switch fails and it is diverted to the Local folder, which you could potentially handle with an extra "Hold job" as you had it in your flow).

[PdFUser5000]

Are you sure you set the folder constraints on the outgoing connections?

Put the connections after Local and Remote on hold and place 5 input jobs in the input folder. The folders should each get 1 job and there should be 3 jobs waiting before "Hold job". Release one of the two connection and see what happens. The released folder should now get a job after the previous one was processed and the number of waiting jobs drops to 2. If that is not the case, then there is something wrong with the settings. I have used this method often enough to know that it works.

Put differently, the number of jobs in Local and Remote should never be higher than 1 (unless the sending to the remote Switch fails and it is diverted to the Local folder, which you could potentially handle with an extra "Hold job" as you had it in your flow).

Everything you described happens so it must be correct. The problem is though that sometimes, when the Local folder is available, Switch still sends the job to Remote. This happens when a single job goes through with no queue. Sometimes it moves to Local, sometimes to remote. I would like the job to always move to Local and only move to remote when local is busy. Or is this how hold job is meant to function, since it's cyclical and there is only one job in the queue?

[freddyp]

"Connection priority"

[PdFUser5000]

"Connection priority"

Thanks!!, works like a charm

[pfischer]

Hello everyone,
this thread is a bit older, but I still have a question that fits the topic.
Freddy mentioned above that the "hold job" element would distribute the jobs in turn if there are several outgoing connections.

We now have the following requirement: we want 9 jobs send to the first outgoing connection and every 10th job to run in a second connection.
I.e. jobs 1-9 go into folder 1 and the 10th job into folder 2. Jobs 11-19 then go back into folder 1 and the 20th job back into folder 2. Is this easy to implement?
Our idea was to provide the hold element with 10 output connections and to merge the outputs 1-9 back into one folder. This means that we end up with 2 target folders ("1-9" and "10" - see the screenshot). Basically this works, but it looks a bit wild and perhaps it could be configured a bit more streamlined.
Do you have any ideas?
Thanks in advance!
Peter

Every 10th job.png (27.6 KiB) Viewed 39836 times

[loicaigon]

What about using "Assemble Jobs" upfront and group them by an arbitrary count of 9.
Once assembled, send them through to the "Hold job" with cyclic.

[pfischer]

I believe that the first problem is that only job packages of 9 are then always packed.

And secondly, all jobs are only forwarded if the number is a multiple of 9. I.e. if, for example, only 12 jobs arrive, 9 are forwarded, but 3 jobs remain in the inbox folder until a further 6 jobs arrive.