#rb Per.Larsson
#jira UE-160942
#rnx
#preflight 6336f2ab5c2225fe5f6c08a5
- Removed the status enum from FPushRequest and replaced it with a class that represents the push result instead.
-- This lets us return more context about problems. At the moment it is only being used to return the filter reason if a payload is filtered out. In the future we could return per payload error messages etc.
- Now that we return more detailed info about the payload push we no longer report payloads that were already in the backend as being pushed which could cause some very misleading stats to be shown.
- Changed the backends to use the new results system.
-- Not all paths were correctly setting the state in the file system backend.
-- The source control backend does not set the result in all paths, which will leave the results in the "pending" state. Which will still mark the payload as not uploaded and so is safe to do.
- Removed the pushing of a single method from IVirtualizationBackend as it was only used in one place internally, so we might as well just change that code to call the batch push overload.
[CL 22279127 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-163834
#rnx
#preflight 632d98e4b4515b7e221654ec
### Virtualization
- At the moment there is no way in the editor UX to change the source control settings for the source control backend, so what ever settings are found in the global environment would be applied when the editor is first run and then saved to a config file under the <saved> directory. From this point onwards those settings would always be used when setting up the source control backend. If the settings were wrong, the first time that the editor was started, then we'd continue to use the incorrect settings even if the global enviroment was fixed. Making the backend work at that point would require that the user know about the local ini file and change the settings there.
- We cannot fix this by just ignoring the ini file as it has been requested that users can add their settings there in case they want to customize things, so we just need to avoid saving to it.
- Added documentation to the source control backends header file on how to set up the ini file if needed.
### Perforce
- FSourceControlInitSettings now accepts EConfigBehavior which describes how we want the source control provider to deal with its settings with regards to the ini file.
- Note that EConfigBehavior is not a bitfield as we do not wish to support only writing to the ini file, there is no point if it cannot be read from.
- When the source control provider is creatred we pass on the into from EConfigBehavior to the FPerforceSourceControlSettings to enable/disable saving and loading.
[CL 22163769 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-164284
#rnx
#preflight 632b182ffc7f1efbdf9638d9
- This is only useful if the entire team uses the same address for the perforce server
- When set, this will override the environment settings for P4PORT as well as override what ever is saved to SourceControlSettings.ini
[CL 22122737 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-163103
#rnx
#preflight 6318989c2b7fe03eb664e9f0
### VirtualizationSystem/VirtualizationManager
- Added an overload of ::Push taking just one FPUshRequest so that people don't have to keep adding MakeArrayView boiler plate when pushing a single request
- Change the order of the last two parameters for the raw ::Push call as this will group all of the payload specific parameters together and leave the target storage type at the end. It is unlikely that anyone is calling the older version but it has been deprecated for safety.
### IVirtualizationBackend
- The none FPushRequest overload of ::PushData is no longer virtual, it just converts the parameters to FPushRequest and calls that overload instead. In this way we now only have one pushing code path in our backends. We could probably look into removing this overload at this point (since the higher level IVirtualizationSystem will now convert all push requests into a FPushRequest form) but it is not considered worth it at the moment when the simple overload covers our needs.
- Removed EPushResult in favour of just returning true/false for the overall operation.If the caller needs a more detailed breakdown then they will have to use an overload that takes an FPushRequest over raw parameters.
-- At the moment FPushRequest does not contain a full breakdown of what happened, so with this submit we are effectively losing the ability to find out if the payload was already in the backends or not, however the batch version of push was already not returning this info so it is not a big loss. Fixing FPushRequest to return a better break down of what happened will be done in UE-160942
- Removed the none batch Push paths from the source control and ddc backends as they already supported batch pushing.
- File backend needed to be converted to supporting batch pushing, which is pretty much the same code as before except we need to iterate over the container of FPushRequests.
-- The backend does not early out on error as it tends to be quite fast. We might want to consider an official policy for the VA system, if we should early out of errors or not.
[CL 21907558 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-161973
#rnx
#preflight 63074424ae13a5a09865a281
- We could try force loading and enabling the plugin at this point but I am not a fan of the idea of overriding the set up of the target in a way that is not immediately obvious to the end user.
- The virtualization module needs to include the 'Projects' module to be able to access the plugin system directly and while editing the file I took the opportunity to clean up how the module is declaring its dependencies.
[CL 21684932 by paul chipchase in ue5-main branch]
#rb Devin.Doucette
#jira UE-161599
#rnx
#preflight 6303c8d65a5d4e4624e7bf52
- There are some use cases that require the VA system to be initialized and configured correctly but would prefer that the backend connections only run if absolutely needed (usually when a payload is pulled or pushed for the first time), this change provides four different ways of doing this:
-- Setting [Core.VirtualizationModule]LazyInitConnections=true in the Engine ini file
-- Setting the define 'UE_VIRTUALIZATION_CONNECTION_LAZY_INIT' to 1 in a programs .target.cs
-- Running with the commandline option -VA-LazyInitConnections
-- Setting the cvar 'VA.LazyInitConnections' to 1 (only works if it is set before the VA system is initialized, changing it mid editor via the console does nothing)
--- Note that after the config file, each setting there only opts into lazy initializing the connections, setting the cvar to 0 for example will not prevent the cmdline from opting in etc.
- In the future we will allow the connection code to run async, so the latency can be hidden behind the editor loading, but for the current use case we are taking the minimal approach.
-- This means we only support the backend being in 3 states. No connection has been made yet, the connection is broken and the connection is working.
-- To keep things simple we only record if we have attempted to connect the backends or not. We don't check individual backends nor do we try to reconnect failed ones etc. This is all scheduled for a future work item.
- If the connections are not initialized when the VA system is, we wait until the first time someone calls one of the virtualization methods that will actually use a connection: Push/Pull/Query
-- We try connecting all of the backends at once, even if they won't be used in the call to keep things simple.
- Only the source control backend makes use of the connection system. The horde storage (http) backend could take advantage too, but it is currently unused and most likely going to just be deleted so there seemed little point updating it.
- If we try to run an operation on an unconnected backend we only log to verbose. This is to maintain existing behaviour where a failed backend would not be mounted at all. This logging will likely be revisited in a future work item.
[CL 21511855 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-161296
#rnx
#preflight 62fe3ce73d3fb466b229bcc0
- There are some usecases that require the VA system to initialize the first time it is accessed (usually the first time we attempt to pull a virtualized payload) rather than be initialized in the program start up. This change provides three different methods to achieve this:
-- Setting the define 'UE_VIRTUALIZATION_SYSTEM_LAZY_INIT' to 1 in a programs .target.cs
-- Setting [Core.ContentVirtualization]LazyInit=true in the Engine ini file
-- Running with the commandline option -VA-LazyInit
- If we detect that the source control backend is being initialized on a background thread we do not try to run the FConnect operation. The backend will still work but this does reduce the potential error checking on initialization. This is done because the FConnect operation currently only works on the main thread and to change this would be a bigger work item than we can schedule at the moment.
- UE::Virtualization::Initialize functions now take a EInitializationFlags enum as a parameter. This enum allows the call to ignore all lazy init settings and force the initialization immediately. This is useful for programs like the Virtualization standalone tool which just needs to start the system when needed.
-- The call to ::Initialize in LaunchEngineLoop passes in None and does not ignore lazy initialization.
-- Calls to ::Initialize in the UnrealVirtualizationTool however all use EInitializationFlags::ForceInitialize and ignore lazy initialization settings.
- Fixed an odd bug in UE::Virtualization::Initialize where the error path (if the config file cannot be found) was using a different start up code path.
- Add asserts when assigning to GVirtualizationSystem to make sure that it is null. This is not 100% safe but should catch some potential threading issues, if any.
- Add an assert after lazy initialization (IVirtualizationSystem::Get) to make sure that GVirtualizationSystem was assigned a valid object.
- Improve how we check for legacy values in [Core.ContentVirtualization]. We now support multiple allowed values.
- Added a way to poll if a VA system has been initialize yet or not, this allows us to avoid initializing a VA system if one has not yet been created and we try to:
-- Dump VA profiling stats after cooking
-- Send VA stats to studio analytics
- Note that currently using lazy init loading will probably cause the VA statistics panel not to work, this will be fixed in future work where we will allow the panel to register for a callback when the system is initialized.
[CL 21467510 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-160954
#rnx
#preflight 62fb9de9086f90bbc4394829
- The problem with the old logic was that we checked all of the payloads initially to see which needed to be submitted but if enough payloads were requested we would split the submission into a number of batches. If a payload was in more than once batch then it would be submitted in the first batch and then fail the MarkForAdd command in the subsequent batches. When we detected this error we would fail the virtualization process.
- In the long term we should probably eliminate duplicate requests at the highest level (FVirtualizationManager) so no backend has to worry about it, but we need this working now so it was easiest to add duplicate checks to the source control backend
-- Note that if another process submits a payload between the status check and the mark for add calls then we will also get errors, but that is true for a lot of our source control code.
- Now when pushing payloads we first iterate over the requests and create a unique list of requests to submit. This eliminates the possibility of duplicate payloads being submitted and failing the MarkForAdd command.
- We now record the push request state for each payload in a map and then only apply it to the original requests if the push succeeds.
-- We still only really have failed/success statuses and nothing recording "already in the backend" which is scheduled to be fixed elsewhere in the backlog.
#ROBOMERGE-AUTHOR: paul.chipchase
#ROBOMERGE-SOURCE: CL 21404920 via CL 21408168 via CL 21408189 via CL 21408222
#ROBOMERGE-BOT: UE5 (Release-Engine-Staging -> Main) (v975-21357124)
[CL 21411940 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-160943, UE-151671
#rnx
#preflight 62fa31bbae3edb54c979e53e
- By checking the state of the payloads first we can avoid loading and writing a payload to disk, just to find that it is already in the depot. If we are very lucky we might find that all of the payloads are already uploaded, in which case we can avoid creating the submission workspace as well.
- The lock to prevent too many concurrent p4 connections does cover a wider area of code than it did before, in theory we could relinquish the lock while creating the payload files for submission, but generally the time it takes to do that is much smaller than the source control operations themselves so there is probably not much point.
- Any request that we find is already uploaded we can just mark as a success. We then create an array of pointers to the requests that were not found and operate on that for the rest of submission.
- Note that we also split up the payload existence checks into batches with the same max size limit of submission, although checking if a payload exists should be much less taxing on the system I'd rather not risk overloading the perforce server where possible.
#robomerge FNMain
[CL 21386511 by paul chipchase in ue5-main branch]
#rb trivial
#jira UE-160083
#rnx
#preflight 62f64714185b21882a875427
- When this backend was written, it shared the source control provider interface with the rest of the editor, so if we were unable to restore the workspace it would likely break the editor in interesting and hard to diagnose ways.
- Given the scope for problems and the low change of the operatation I opted to make the error fatal.
- Now that the backend owned it's own source control provider interface, the worst thing that can happen is that future virtualization attempts will also fail, but the user will not be left with a non functioning editor. With this in mind there is no longer a reason to consider this a fatal failure.
[CL 21353746 by paul chipchase in ue5-main branch]
#rb trivial
#jira UE-156189
#rnx
#preflight 62f5f1a01e61d1ba0e6c893b
- Note that the older name 'SourceControl' will continue to work but will log a warning if used to notify the user to update their config files.
[CL 21352223 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#rnx
#jira UE-160619
#preflight 62f37c86b66d5d93133d67e3
- Based on work from Jess.Kube
### Virtualization
- The source control backend now takes an optional config option "ClientStream" which takes the name of the client stream to use.
- If a client stream is set then workspaces created for payload submission will use that and not provide a client-view mapping.
#ushell-cherrypick of 21195584 by Jess.Kube
### PerforceSourceControl
- Allow FCreateWorkspace to create workspaces with streams as well as classic workspaces cia FCreateWorkspace::SetStream.
- Add a method FCreateWorkspace::ClearClientViewMappings which will clear any client view mappings already added to the operation.
- If we detect that a FCreateWorkspace operation has both a stream set and client view mappings set then FPerforceCreateWorkspaceWorker will return an error. Perforce will allow us to creat a client spec with both entries, but will default to using the stream. Technically we could allow this too but it might cause unexpected behaviour to the caller. It is better to give a clear error and fail the workspace creation.
[CL 21316756 by paul chipchase in ue5-main branch]
#rb trivial
#rnx
#jira UE-159992
#preflight 62d7c9eb110da0f944670ecb
- When the depot path is parsed from the config file we now add the trailing '/' if it is not already there, this allows us to make the input more forgiving without changing the existing code.
- Removed FString::Printf when creating the path for payload_metainfo in favour of using WriteToString, this doesn't add much but gets rid of the %s.
- Reformatted some messages that had odd tab alignment.
[CL 21176124 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-158771
#rnx
#preflight 62c6c8f2f5590c326d77b0f3
- This is a temp fix to an internal problem, where we have a group of people unable to access the source control payload storage system due to permissions but do not actually need to access it at the moment.
- By supressing the connection failure pop up we can allow them to keep working undistrubed.
- Improving how we handle error notifications like this is already on the VA roadmap, at which point this change can be removed.
#ROBOMERGE-AUTHOR: paul.chipchase
#ROBOMERGE-SOURCE: CL 20982887 via CL 20982891 via CL 20982896
#ROBOMERGE-BOT: UE5 (Release-Engine-Staging -> Main) (v972-20964824)
[CL 20984627 by paul chipchase in ue5-main branch]
#rb trivial
#rnx
#preflight 628c9337a24685c36f50ec2f
- When the code was converted from single batch to support multiple batches I missed the check to return true if all payloads were already uploaded. So once we were operating on multiple batches if a batch was 100% already present in the backend we would return true and not process any further batches.
- Change the return true, to a continue so that future batches are processed.
[CL 20345282 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#jira UE-151000
#rnx
#preflight 627291732f6d177be3c292e3
- When submitting large batches it was found that once a cl description reached a certain length, that p4v would lock up attempting to display it in the history tab, so we want to be able to make smaller submits to keep the description to a managable size.
- In addition it is better for our server set up to make multiple smaller batch submits than insanely large ones.
- The max size of the batch can be set when creating the virtualization graph in the config file.
[CL 20055325 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#rnx
#jira UE-151000
#preflight 627271c83b2ed6f85363f53a
- In the old code when submitting packages for virtualization we would load all local payloads into memory then pass them to the virtualization system to be pushed to persistent storage. This works fine for an average users submits but when resaving large projects we can have submits with thousands of packages containing gigabytes of payloads.
- FPushRequest has been changed to accept either a payload directly (in the form of FCompressedBuffer) or accept a reference to a IPayloadProvider which will return the payload when asked.
- Most of the existing backends make no use of the batched pushing feature and only push one payload at a time. The default implementation of this will be to ask each request one at a time for the payload and then pass it to the backend for a single push. If the IPayloadProvider is being used we will therefor only ever have one payload loaded at a time avoiding memory spikes.
- The source control backend currently does implement the batched pushing feature. This backend writes each payload to disk before submitting them to perforce, so all we have to do is make sure that request one payload, write it to disk, then discard the payload to avoid memory spikes.
- This change required that FPushRequest make it's members private and provide accessors to them, as the caller shouldn't need to care if the payload is loaded or comes via the provider.
- NOTE that this implementation is less efficient if we have packages containing many different payloads as we will end up opening the package, parsing the trailer then load a payload many times over, where as the original code would load all payloads from the file in one go. In practice, given the overhead of the source control backend this didn't make a huge difference and is probably not worth the effort to optimize at this point.
[CL 20040609 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#rnx
#jira UE-147948
#preflight 6271209be16e280be60b5a62
- In the worst case scenario (where a machine can only access the source control backend) we will want to limit the potential number of concurrent connections that they can make.
- In testing, not limiting the connections could cause the occasional connection failure and I imagine having a full team running in the worse case scenario would quickly overwhelm the server.
- Currently we default to 8 connections (the limit allowed in UGS)
- The semaphore is a quick and dirty implementation similar to std::counting_semaphore. I am not spending much time on it as the worst case scenario is not expected to be common and if it does occur, the system is going to be painfully slow no matter how good the thread throttling.
### Future work
- This work could be taken a lot further by either
-- a) reusing connections rather than creating one for each pull, which will require changes to our source control api
-- b) gathering requests from many requesting threads once the limit has been reached and then sending the requests in batches
--c) reserve one connection for the game thread to prevent the user waiting?
[CL 20024594 by paul chipchase in ue5-main branch]
#rb PJ.Kack
#jira UE-148223
#rnx
#preflight 6260f6e4360b45c32a84fae8
- The commandline -VA-MissBackends has been extended to a console command 'VA.MissBackends' that can be used at runtime to disable specific backends or reset the list and enable all of them again.
-- Entering VA.MissBackends without any args will print help documentation to the console/output log and show the user the names of the backends that they can disable.
- Technically we don't need the command line parsing anymore and users could use -dpcvars=VA.MissBackends=X instead but since we already have the code we might as well leave it untouched.
- Disabling backends from pulling payloads for debug purposes has been split off to it's own system stored as a
- IVirtualizationBackend::EOperations has been turned into bit flags, so we no longer need the 'both' entry
- IVirtualizationBackendnow has two EOperations members, one to record the operations that the backend supports and another that allows the operations to be disabled for debugging purposes.
- IVirtualizationBackend now has one method to poll if an operation is supported ::IsOperationSupported, which should not change over the lift time of the backend. In addition there is also a method for disabling/enabling a backend for debugging purposes as well as polling the debug state.
[CL 19845175 by paul chipchase in ue5-main branch]
#rb Per.Larsson, Sebastian.Nordgren
#jira UE-147948
#preflight 6256ba2f647ad886b380d007
- In some cases people are getting minor connection failures to the perforce server which causes the p4 print command to fail. In these cases it would be better to retry the command rather than failing to get the payload entirely
- Currently the default will be 2 attempts to pull a payload with a 100ms wait between them, this can be changed when setting up the backend graph as optional config values "RetryCount=" and "RetryWaitTime="
-- Note that RetryWaitTime is in ms
- Moved the parsing of the command line to it's own method in the backend.
[CL 19752733 by paul chipchase in ue5-main branch]
#rb Sebastian.Nordgren
#rnx
#preflight 622f645a19287beb2c359176
- Use the SetRequireDirPathEndWithSeperator option with the FUpdateStatus operation as we are checking depot file paths. If we do not use this option they can be seen as network paths and will be checked against the file system which can take some time to time out.
- Wrapped this up in a utility function so that it can be more easily updated in the future.
[CL 19373558 by paul chipchase in ue5-main branch]
#rb Per.Larsson
#rnx
#jira UE-136480
#preflight 62289b59f4469cadac10eb91
- The backend can now connect to source control without forcing the rest of the editor to connect to source control.
- If the user disconnects the editor from source control, the backend will still be able to function using it's own connection.
-When the backend needs to change the current workspace in order to submit payloads we no longer invalidate the file cache used by the editor and so the user will no longer have to manually refresh files in the content browsers to restore the source control status icon.
[CL 19318970 by paul chipchase in ue5-main branch]
