Here’s my RC6 feedback/ramblings. Disclaimer - [random thoughts/ideas/things I saw after a busy day at work / ignore as applicable!]
DJV plugin has been dropped? Any other plugins dropped in v6?
How far away are we from multiple Pulses running in a failover mode?
PDF manual - page 15 - “This is just a placeholder. The wiki extractor for Deadline will replace this page with the general licensing page.” - ???
• page 15 - " Users of CentOS 5 and 6 have to build Mono from source. You can download our Mono Build Script to automate this process."
would be good to actually provide links here to download the build script
Monitor user options - missing features?
• line numbers in Report Viewers
• Maintain List Selection after refresh?
• slave ping timeout
• Use 24hr clock
• net send is gone same as Growl on the Win platform?
view grid lines in monitor?
monitor > Configure Plugins > CommandLine / CommandScript / Draft > text is cut-off? “There are no configurable settings for this plug”…“plugin”?
Missing features?
Tools > “Undelete Jobs”, “Check for Orphaned Limit Stubs”, “Empty Trash Bin”, “Suspend ALL Active Jobs”, “Resume Previously Suspended Jobs”
Repository Options - Missing Features?
Where is Pulse “Repository Disk Space Monitoring” warning if disk space < x MB & email notification for low disk space warning?
Thinkbox remote error reporting service?
email notification of corrupted jobs?
Job Properties - Missing Features?
job properties of a 3dsMax job > 3dsmax settings > text is getting cut off - “Ignore Missing UVW Corrdinat…”
job properties > right-click > expand ALL / collapse ALL functionality lost?
example “Script UI Example” script missing from scripts menu
install “Install Integrated Submission Scripts” script missing from scripts menu
We trimmed down on some plugins. We had converted the majority of the plugins around the start of the beta, and since no one made any mention of the few we left out, we just left them out (this is the main reason why DJV didn’t make it over). If there are ones people want to see back, we could always look at revisiting them for a future release.
Here’s the list of the plugins we’ve dropped:
CleanerXL
DJV
FrameFixer
FryRender
Gelato
Indigo
IronPython
ParticleIllusion
POVRay
QuicktimeFusion
SequencePublisher
VNS
xNormal
Pulse is no longer used as a proxy, and thus isn’t required to scale up. It’s only necessary if you want to use some of the advanced features like Power Management, Stats Collecting, and the Web Service. Since it’s not being used as a proxy anymore, I don’t think this is even necessary.
Thanks for catching that Licensing issue in the PDF. We’ll get that fixed! Regarding the links though, currently none of the external links in our docs make it into the PDF.
Line numbers didn’t make it over. We just didn’t have time to dig into how to do it with Qt. It’s currently on the wishlist for 6.1.
The list selection is always maintained with the auto refresh system, so this isn’t necessary.
Pinging is done in a background thread with a timeout of 1000 ms. We figured we wouldn’t make this configurable unless it was necessary.
24 hour clock is always used now.
Net send and Growl are gone, but you can always write an event plugin to replace them.
Grid lines are now always off. No one complained during the beta, so we just stuck with that.
Thanks! We’ll get this fixed.
Missing by design. The check for orphaned limit stubs and emptying the trash bin are replaced by the housecleaning option (which does everything). The Suspend All system was left out because it had some issues. If people want it back, we can take another look at it.
We took out the repository disk space check since it only worked if you were running Pulse on the same machine as the repo. We didn’t like that it only worked in some cases.
Remote error reporting has been discontinued. To be honest, 99% of the reports we got were file access errors, making it difficult to find real problems. If I recall correctly, the system hasn’t produced a helpful message that resulted in a bug fix for a long long time.
With the new database, jobs should no longer become corrupted, unless there is an actual bug in Deadline. Corrupted jobs are no longer considered a “feature”.
Text cut off has always been an issue with the plugin specific settings.
The expand/collapse all option was an oversight though. We’ll see if we can get this into 6. If not, then 6.1.
Those didn’t make it over either. The Script UI example was left out mainly because the scripts that we do ship with Deadline provide great examples.
The integrated script one didn’t make it over due to time constraints. It could always come back if there is demand though.
Not that I have any vested interest in DJV or any other plugin being continued in v6. Just thought that only time will tell when v6 goes out to the masses, as to which plugins are potentially missing. I just don’t like the idea of a potential new customer discounting Deadline just because it was missing plugin X and plugin Y, when in fact they were built years ago and just need a simple port to make them operational in v6. However, sometimes, there comes a time to trim the dead wood.
I’m going to disagree with you on this one. Our renderfarm turns OFF to save electricity. If Pulse goes down, I have jobs hitting my queue and no ability to automatically turn the machines back on again. The WebService is also important to allow our users to check all is well on the farm and they don’t need to travel into the studio to resolve issues when close to a deadline. I’m really just looking for a heartbeat failover Pulse setup. Another machine to pick up the work, if the master flops on me. Finally, just to be pedantic, the new monitor tab says “Pulses”. That’s plural!
line numbers would be good in a future release.
I get a sense other users will be requesting a netsend/growl event plugin upon release. Time will tell
Grid-Lines…damm…I should have noticed that earlier…Again…time will tell.
Shame about “repository disk space check”. In Deadline v5, I have Pulse running on a separate machine to the repository (for performance) and although I haven’t checked for a long time, I was under the impression the disk space check works just fine. It would be a good feature to get working again in v6.
“Remote error reporting” - out of scope for now, but having more of a syslogger (logstash) coming out of Deadline upon an end-users agreement, tunnelling into a cloud storage for you guys and then analysed via something like Kibana/Graylog2, to find the highest error counts/messages might be quite interesting to explore. I really like the new “console” window. You could almost re-direct the console stacktrace/messages to the above setup and pretty much have a remote stacktrace coming from each of your client securely, ready for analysis…
Excellent
ScriptUI Example / Install Integrated Scripts - Understood. They all just help to save time/sell the product as easy/user friendly to setup
I miss the Undelete jobs option. I don’t understand why people think it is OK to delete problem jobs that they know are being trouble shot at the time, but they do. If deadline 6 doesn’t delete the jobs immediately (does it wait until a house cleaning?), how do we get them back without editing the database?
Also, the ‘Find Task in Task list’ from the slave pane is also sorely missed.
There is a level of disbelief remaining that the Remote Desktop trigger no longer has a hotkey, and takes two steps (though it takes one step here now - I haven’t looked at whether we can add hotkeys).
Hi,
I regularly use the “Undelete jobs” option to help artists out. V. important!
Again, “Find Task in Task list” is super important. Use this all the time together with its reverse brother “Find Slave in Slave List”
Mike
Good point about the Pulse failover. We can add this to the wish list. A very quick way to implement something like this would be for other pulse’s to ping the pulse that’s configured in the repository options to make sure it’s responding. After a few failed attempts, one of the other pulses would just autoconfigure the repository options to make it the primary pulse. I think the only situation this doesn’t cover is the web service (since the server address will now be different). However, if you had some sort of proxy server running for your web service that could automatically redirect to the primary pulse, that would solve this.
We’ve added Undelete Jobs to the todo list for 6.1, and keyboard shortcuts for scripts is already on the todo list for 6.1.
The reason why Find Task in Task List and Find Slave in Slave List didn’t make it over is that at the time we couldn’t figure out a way to do this properly now that you can have multiple slave and task lists open at the same time. Thinking about it now though, maybe the solution would be to have these options spawn a new panel with the appropriate row selected. This should work fine for finding a slave from the task list, but finding the task from the slave list becomes interesting because the task list contents are dependent on the selected job. Now if we replaced it with Find Job in Job List, then we could just spawn a new job list with the job selected. What do you guys think? If you like it, we can add this to the todo list for 6.1.
On the Pulse failover, a PING to the current Pulse machine wouldn’t be enough as the actual machine might respond OK, but the Pulse at an application level could be screwed. It would be better to query the actual Pulse application as well as PING the basic machine. (In v5 I have seen Pulse application hang, but the machine was fine. Like when there were too many *.tmp files being created for Pulse to purge. This was fixed in v5.2 BTW.)
So, 2 step failover/heartbeat check process:
basic PING to actual Pulse machine as currently configured in repository AND if Pulse operational (enabled)
query Pulse application if stage 1 above is successful, check if Pulse at application level is responsive. Could use equivalent “WebService” type response to confirm all is well.
As you said, WebService redundancy is dependent on individual Studio setup. Proxy would work.
On the Task / Slave List selection issue for multiple panels, I guess it’s not possible to just highlight the slaves or tasks in ALL the panels and let the user decipher which is relevant? Or are you saying, its tricky to achieve this UI result via QT? Not sure about spawning another list…
On a separate note, whenever on OSX I spawn a slave console log, Pulse log, etc, the panel always appears behind the monitor application that I executed the command to spawn from. Really annoying as I have to go hunting for it!
Sorry, by “ping” I meant to talk directly to pulse, not just ping its machine to see if it was online.
We had thought about highlighting the task or slave in all panels, but what do you do if it’s currently filtered out? I guess the bigger question is WHY do you use this feature? If it’s to quickly connect to the task’s slave, would spawning the slave panel give you the result you’re looking for?
Thanks for the note about log panel. We’ve logged this as a bug targeting 6.1.
Use of the highlight task or slave in its list, is primarily to identify what a certain machine is up to at that exact moment in time OR how long a particular task of a job has been running on a certain machine.
Both use cases, are generally to decide if a task or machine can be re-queued or taken off-line. ie: “That task has been running for ages and is at 99%. Let it finish before taking that machine off-line for maintenance.” OR “Why is that particular task taking so long…Ah…it’s because it got picked up by machine X which is really slow”
So it sounds like Task -> Slave would work fine by spawning a new Slave panel, right?
For Slave -> Task, what if we just stored all the task information in columns in the Slave list? We already have Task frames, progress, and status, and adding another for Task render time should give you everything you need.
If you are suggesting not having the Find Task in Task List selecting the job in the job list, then I’m not for that at all: I don’t actually use this option to find the task in the task list, I use it to find the job in the job list.
Spawning a new job pane would be fine, or reusing where appropriate if filtering allows.
Thanks Grant! So it sounds like the following could please everyone:
In the Task list, have a “Find Slave” option, which spawns a new list with the Slave selected.
In the Slave list, have a “Find Job” option, which spawns a new list with the Job selected.
In the Slave list, show the task render time in a new column, so that essentially all task information you would see in the task list for a rendering task would be shown in the slave list.
If this sounds good, we’ll put it on the todo list for 6.1.
Hi,
Yep, this sounds fine. The only thing I would add, is if possible, displaying the filtering result within the current panels without having to spawn another new panel would be slick, but I do understand that this might be a bit too tricky!
Thanks,
Mike
Yeah i would also favour reusing existing panels if the filters allow. There is so much window popping / automatic reorganization / drag’n’droppage going on already, its a bit overwhelming.
If we did stick with existing panels, what do we do if there is more than one instance of that panel already visible? What do we do if the job is currently filtered out? We’re just not sure what the “correct” course of action is to take in these cases.
We’ve been playing around with using a new panel, and using filters on that new panel to only show the relevant job(s) or slave(s). What’s cool is that you can select multiple tasks and the slave panel that pops up will contain just the relevant slaves, or you can select multiple slaves and the job panel that pops up will contain just the relevant jobs.
