Disabled by default. Currently can only be enabled with console "debug lludp oqre start" command, though this can be started and stopped whilst simulator is running.
When a connection requires packet queue refill processing (used to populate queues with entity updates, entity prop updates and image queue updates), this is done via Threadpool requests.
However, with a very high number of connections (e.g. 100 root + 300 child) a very large number of simultaneous requests may be causing performance issues.
This commit adds an experimental engine for processing these requests from a queue with a persistent thread instead.
Unlike inbound processing, there are no network requests in this processing that might hold the thread up for a long time.
Early implementation - currently only one thread which may (or may not) get overloaded with requests. Added for testing purposes.
This records how many packets were indicated to be resends by clients
Not 100% reliable since clients can lie about resends, but usually would indicate if clients are not receiving UDP acks at all or in a manner they consider timely.
This allows one to monitor the total number of messages resent to clients over time.
A constantly increasing stat may indicate a general server network or overloading issue if a fairly high proportion of packets sent
A smaller constantly increasing stat may indicate a problem with a particular client-server connection, would need to check "show queues" in this case.
This check is necessary to close a race condition where the CompleteAgentMovement processing could proceed when the UseCircuitCode thread had added the client to the client manager but before the ScenePresence had registered to process the CompleteAgentMovement message.
This is most probably why the message appeared to get lost on a proportion of entity transfers.
A better long term solution may be to set the IClientAPI.SceneAgent property before the client is added to the manager.
This is to deal with one aspect of http://opensimulator.org/mantis/view.php?id=6755
With the V2 teleport arrangements, viewers appear to send the single UseCircuitCode and CompleteAgentMovement packets immediately after each other
Possibly, on occasion a poor network might drop the initial UseCircuitCode packet and by the time it retries, the CompleteAgementMovement has timed out and the teleport fails.
There's no apparant harm in doubling the wait time (most times only one wait will be performed) so trying this.
Add more information on which endpoint sent the packet when we have to wait and if we end up dropping the packet
Only check if the client is active - other checks are redundant since they can only failed if IsActve = false
Used to detect if a simulator is receiving significant junk UDP
Decimates the number of packets between which a warning is logged and prints the IP source of the last malformed packet when logging
The root cause was that v2 was only closing neighbour agents if the root connection also needed a close.
However, fixing this requires the neighbour regions also detect when they should not close due to re-teleports re-establishing the child connection.
This involves restructuring the code to introduce a scene presence state machine that can serialize the different add and remove client calls that are now possible with the late close of the
This commit appears to fix these issues and improve teleport, but still has holes on at least quick reteleporting (and possibly occasionally on ordinary teleports).
Also, has not been completely tested yet in scenarios where regions are running on different simulators
What I believe is happening is that on initial terrain send, this is done one packet at a time.
With WaitOne, the outbound loop has enough time to loop and wait again after the first packet before the second, leading to a slower send.
This approach instead does not wait if a packet was just sent but instead loops again, which appears to lead to a quicker send without losing the cpu benefit of not continually looping when there is no outbound data.
This reflects the actual use of this stat - it hasn't recorded general exceptions for some time.
Make the sim extra stats collector draw the data from the stats manager rather than maintaing this data itself.
In this new protocol, and as committed before, the viewer is not sent EnableSimulator/EstablishChildCommunication for the destination. Instead, it is sent TeleportFinish directly. TeleportFinish, in turn, makes the viewer send a UserCircuitCode packet followed by CompleteMovementIntoRegion packet. These 2 packets tend to occur one after the other almost immediately to the point that when CMIR arrives the client is not even connected yet and that packet is ignored (there might have been some race conditions here before); then the viewer sends CMIR again within 5-8 secs. But the delay between them may be higher in busier regions, which may lead to race conditions.
This commit improves the process so there are are no race conditions at the destination. CompleteMovement (triggered by the viewer) waits until Update has been sent from the origin. Update, in turn, waits until there is a *root* scene presence -- so making sure CompleteMovement has run MakeRoot. In other words, there are two threadlets at the destination, one from the viewer and one from the origin region, waiting for each other to do the right thing. That makes it safe to close the agent at the origin upon return of the Update call without having to wait for callback, because we are absolutely sure that the viewer knows it is in th new region.
Note also that in the V1 protocol, the destination was getting UseCircuitCode from the viewer twice -- once on EstablishAgentCommunication and then again on TeleportFinish. The second UCC was being ignored, but it shows how we were not following the expected steps...
If this consistently increases then this is a problem since it means the simulator is receiving more requests than it can distribute to other parts of the code.
If we're not receiving packets with multiple threads (m_asyncPacketHandling) then this is critical since it will limit the number of incoming UDP requests that the region can handle and affects packet loss.
If m_asyncPacketHandling then this is less critical though a long process will increase the scope for threads to race.
This is an experimental stat which may be changed.
AgentUpdate packet. This fixes the problem with vehicles not moving forward
after the first up-arrow.
Code to fix a potential exception when using different IClientAPIs.
Enabling this will stop anybody from moving on a sim, though all other updates should be unaffected.
Appears to make some cpu difference on very basic testing with a static standing avatar (though not all that much).
Need to see the results with much higher av numbers.
This appears to improve cpu usage since launching a new thread is more expensive than performing a small amount of inline logic.
However, needs testing at scale.
Justin, if you read this, there's a long story here. Some time ago you placed SendInitialDataToMe at the very beginning of client creation (in LLUDPServer). That is problematic, as we discovered relatively recently: on TPs, as soon as the client starts getting data from child agents, it starts requesting resources back *from the simulator where its root agent is*. We found this to be the problem behind meshes missing on HG TPs (because the viewer was requesting the meshes of the receiving sim from the departing grid). But this affects much more than meshes and HG TPs. It may also explain cloud avatars after a local TP: baked textures are only stored in the simulator, so if a child agent receives a UUID of a baked texture in the destination sim and requests that texture from the departing sim where the root agent is, it will fail to get that texture.
Bottom line: we need to delay sending the new simulator data to the viewer until we are absolutely sure that the viewer knows that its main agent is in a new sim. Hence, moving it to CompleteMovement.
Now I am trying to tune the initial rez delay that we all experience in the CC. I think that when I fixed the issue described above, I may have moved SendInitialDataToMe to much later than it should be, so now I'm moving to earlier in CompleteMovement.
reception thread to sleep for 30ms if the number of available user worker
threads got low. It doesn't look like any of the UDP packet types are
marked async so this check is 1) unnecessary and 2) really crazy since
it stops up the reception thread under heavy load without any indication.
Justin Clark-Casey (justincc)