Just happened right before 12:13 AM. - Game H

That isn't a lot of info.

There are lots of possible reasons you might get a pause. But I'd be concerned if the pause was more than, say, 5 seconds, and especially if it's happening to everyone consistently. That's why this has been so hard to track. Pauses can be caused by spikes in CPU usage, even by others sharing the same server machine. It can also be caused by a temporary slowdown of your network connection. I see this now and then. It can be caused by a player running an intensive script, especially with all of this added debug overhead making the game a fraction of the efficiency that it will run at when released.

If you've noticed a decrease in pauses, or in the length of the pauses, then I'd see that as progress. I don't expect pauses to go away entirely. If I could guarantee that the pause is an internal TW issue, then sure, I'd want to fix it. But I have no way of knowing that at this point. Do you feel that what you're experiencing is different than the kinds of pauses that you have always experienced?

BTW, the best way for me to diagnose the cause of a pause is for you to drop carrier or relog during the pause so TW kicks your session offline and generates a detailed report. Among other things, I'll know exactly what the game was doing during the pause, as well as whether or not it was able to send and receive IO.

I was able to dig up some useful data from the logs based on your report. I see that there was a pause of almost a minute and a half between outputs at 12:11:58 and 12:13:23. On the input log, I see that there was a "Q" sent at 12:11:58, a keepalive at 12:12:08, then another at 12:12:38, then a gap before a block of input reached TWGS at 12:13:23. This gap is about 45 seconds. Because this data is logged at the moment TWGS reads it off the socket, I know that either TWGS lagged for 45 seconds, or there was a 45 second lag on the socket itself, after which a block of q^mq^mq^m<keepalive>^m was received. Now, I didn't look at every active player, but the first other log I looked at, Xanos, showed high activity during the gap between 12:12:38 and 12:13:23, during which time he processed on average about 40 characters per second. So that tells me that TWGS itself was actively processing socket events during that gap. Based on this data, either there was too much activity and your input was bogged down, or you had a network slowdown. It's clear that the lag that you experienced was external to TWGS, that TWGS did not receive a socket event during that pause, while it received other socket events for other players.

Based on this, I'm inclined to conclude that the lag was unrelated to TWGS and especially to TW. I wish this was more cut and dried, but tracking the cause of intermittent lagging on a server is extremely difficult.

There were only 3 of us logged on, 2 at the keys. The relog was T.R. and Xanos was moving fuel, I was simply lifting off a planet when the lag happened.

What more can I do to help on this end reporting these occurances?

This is all we can do at this point. If I know when a lag happens, I can check the logs to see what was happening at that time and decide if it's likely the lag was caused by TWGS or TW or was external to TWGS and TW. Unless I see any obvious cases where TWGS or TW was locked up for a period of time, as opposed to waiting for input on the socket, I can't do anything to change it.

If people are moving a lot of product in a sector and you lift, your telnet will lag (especially SWATH). Not sure if that is what you were seeing but I always tried to sit at the planet prompt (not citadel) when buying down or moving a lot of product. Were TR and X seeing the lag as well or only you? It was always random on what person would be affected but it seems the movement is buffered so the person sitting/lifting seems to notice it the most.

I was AFK but will be keeping my eyes open for pauses when I am.

I have noticed pauses on non-beta games as well. If I go into a sector where someone is running a high-speed product moving script, or colonizing script, Swath will bog down from the rapid messages. Maybe we can get Stein to look into this issue, since it seems to be only Swath that does it. (I don't see the same issue with TWAR, but I still prefer MUCH Swath.)

In Swath the setting to locate is under Swath Preferences/Miscellaneous

at the bottom High text speed limit:

I have mine set to 7000 chars/sec

It makes the screen look choppy as if it were lagging.
But the advantage is that I don't spike my cpu because of the spam
caused by someone moving product fast.

I also will hide at planet prompt for ck buydowns just to prevent the server from having to send data to each player in sector or citadel.

But if you keep a copy of telnet open , you can see the real speed of data sent to you , regardless what Swath is showing.

I was not in the same sector as Xanos, but thanks for the thought.

Crucher - no problem I was just unclear on what all the setup was of everyone.

Vid - I have played with that setting and it still does the same if someone is doing a buydown or an action that sends a lot to buffer. Might be more noticeable on a server that has a high CPC. I do know if you are using wine in linux running swath it will lag to the point of unplayable but most people are not using that setup.

Very soon I'm going to pull all of the debug code and run it in a fully optimized mode and then we'll see if the lags improve (suggesting it's related to spikes in use). Right now it's not uncommon to see a spike up to like 80% CPU when a script is running, but when fully optimized, no action in the game will generate more than 5% CPU load on this server.

10:42:32 AM Fri May 26, 2023

I don't know what the others were doing, but I was not running a script at the time.


11:12:24 AM Fri May 26, 2023

This time I was running a buydown script.

Any script that moves product to/from planets or ports is faster than it used to be in the old version, and uses more CPU on the players server, and you would think would use more CPU on the game server also. Compared to the timings of the rest of the beta, I think that part is a bit too fast and needs toned down just a hair.
One suggestion would be to have a small ms delay IF the player is NOT negotiating port buy downs.

I need to see a bit more corroboration of this issue, because the reports all seem to be coming from the same people. Since every indication I've seen suggests that the pauses are external to TWGS, ie, TWGS is processing data quickly once it receives it, but there are occasional pauses between the arrival of input, and the same people are consistently reporting the issue while others have not, I have to conclude that it's beyond my control. I've spent two weeks tracking these pauses and just haven't found a shred of evidence that TWGS itself or TW itself is lagging in any way during these pauses.

If everyone is experiencing these problems but just not reporting them, then I would consider the issue still open. But even then, it's possible it's related to this particular server and not the software. I think once the server is out there and more people are using it, I'll get a better picture as to whether this is in fact a bug that I can address.