Darn, you have been logged out of SecondLife.

[Profaitchikenz Haiku]

All because you wanted to TP from region F to region G, and, well, we sort of lost track of things, so when you log back in you're going to be back at region A, but with none of the floaters like inventory and people open that you had set up just how you like them, and no chat history either.

OK, a tad sarcastic I know, but something is puzzling me. If I can TP from A to B to C to D to E to F without a problem, what went wrong with G and why was it unrecoverable?

When I log back in to A and restore my floaters, then TP to G, it succeeds. If I go back to F and try to TP to G, it succeeds. The problem, whatever it is, seems sporadic.

I would far prefer the "Unable to TP you right now..." treatment because I can at least choose to go somewhere else and I still have all the viewer set up as I want it.

My question, or two of them, is/are therefore

a) Why did region F pass across all of the details to region G and forget about me when it seems that region G wasn't ready/able to accept me?

b) why can't the state of the inventory and people floaters and any open notecards be stored in my cache and reloaded after such an episode?

 

It's occurring sporadically in all viewers, and not always the same regions, so it isn't a main-server to RC server issue.

[Wulfie Reanimator]

3 minutes ago, Profaitchikenz Haiku said:

a) Why did region F pass across all of the details to region G and forget about me when it seems that region G wasn't ready/able to accept me?

When you teleport from one region to the next, the region you're leaving is basically throwing you out and yelling "Catch!" to your destination. If your destination misses that event, your viewer is no longer connected to any region and you get logged out. There's no reconnect feature and nobody's bothered to make one since "it's just easier to restart the program."

The reason why you end up all the way back to region A in your example is because the avatar's state or "last location" is only saved during a proper logout process. When you fall into the void, there's just nothing to save you.

 

3 minutes ago, Profaitchikenz Haiku said:

b) why can't the state of the inventory and people floaters and any open notecards be stored in my cache and reloaded after such an episode?

They could, it just hasn't been done yet.

[Monty Linden]

47 minutes ago, Profaitchikenz Haiku said:

a) Why did region F pass across all of the details to region G and forget about me when it seems that region G wasn't ready/able to accept me?

Don't know but part of the reason may be hiding in your SecondLife.log file.

[Profaitchikenz Haiku]

48 minutes ago, Wulfie Reanimator said:

They could, it just hasn't been done yet.

I know. I was just venting.  expressing the wish that this problem might get a little more attention once they've finished making everybody look so beautiful.

[Profaitchikenz Haiku]

13 minutes ago, Monty Linden said:

Don't know but part of the reason may be hiding in your SecondLife.log file.

I have been studying these diligently, and it's scary just how many errors and timeouts get reported in there. but these are all associated with good TPs .

At present I don't see any extra lines in the run up to the failed TP that hint at any problem, but I am now going to see if there are any characteristic lines that are absent in the run up to a failed TP. Perhaps there's something like I am trying to move off whilst the cache is still being loaded... something drawing resources from the required exchange of data?

If there is any guide to understanding the log files I would be glad to see it. (ETA I know, you're going to say "So would we")

Edited March 20, 2023 by Profaitchikenz Haiku

[Monty Linden]

25 minutes ago, Profaitchikenz Haiku said:

If there is any guide to understanding the log files I would be glad to see it. (ETA I know, you're going to say "So would we")

Hahaha.  No, the answer is even more hilarious.  LL was never interested in that so the one bit of writing I did on decoding the log file was actually published in Firestorm's Jira.  (Found it:  https://jira.firestormviewer.org/browse/PHOE-2707 - damn, that was a long time ago.)

The log file *is* ugly but looking for missing events is a promising technique.  There is a TP failure mode around missing 'TeleportFinish' messages and/or responding to them slowly.  Something to look for...

[Profaitchikenz Haiku]

Interesting. The first 8 lines of a successful and an unsuccessful TP in the log are identical (beginning with a teleportRequestViaLandMark and ending with a llinventorymodelbackground fetch)

Then, they diverge completely.

A successful TP continues with a cleanup message followed by addCircuitData and some setting up of the new connection.

A failed TP shows four messages about memory allocations, six lines about asset storage details, some viewer FPS stats, a mesh download quiescent message, and then the shut down summary. It's obviously made it's mind up right after that inventorymodellbackgroundfetch that it isn't going anywhere.

Both logs show this is all happening within the same second, so it's not a lengthy timeout, if a timeout at all, it's some sort of non-logged exchange that hasn't happened.

I hate to think what the logs would look like if I had full debug output turned on.

I'll have to learn to remember to learn to live with it.

 

ETA (I had a look at that Firestorm Jira, most of the issues there seem to be bake fails and login hangups, shame they didn't have some TP-fails too)

 

Edited March 20, 2023 by Profaitchikenz Haiku
Accuracy

[Wulfie Reanimator]

22 minutes ago, Profaitchikenz Haiku said:

It's obviously made it's mind up right after that inventorymodellbackgroundfetch that it isn't going anywhere.

Both logs show this is all happening within the same second, so it's not a lengthy timeout, if a timeout at all, it's some sort of non-logged exchange that hasn't happened.

I hate to think what the logs would look like if I had full debug output turned on.

That's the thing I was referring to. 🙂 Using a debug build of the viewer doesn't reveal anything out-of-the-box (I've encountered it while working on other things), someone would have to add more logging to that process to see what the viewer is doing, but it probably won't give a full picture of what's happening since we can't see what the sim(s) are doing.

Edited March 20, 2023 by Wulfie Reanimator

[Profaitchikenz Haiku]

4 minutes ago, Wulfie Reanimator said:

Using a debug build of the viewer doesn't reveal anything out-of-the-box, someone would have to add more logging to that process

Yes, it's the old problem of you don't know what to look for until you have to go looking for it.

And ... when you add extra diagnostics, sometimes you chase the problem away. We had problems with memory leaks in a Fortran-90 program that didn't happen when we turned on diagnostics. Little did we know that doing that made unassigned pointer values default to a safe value .

Edited March 20, 2023 by Profaitchikenz Haiku

[Whirly Fizzle]

JIRA issue about this: BUG-41660 - Region handover during teleport unreliable

Quote

Currently, the handover process between two regions of the avatar during teleports is completely unreliable. The current region sends the circuit info of the new region to the avatar and then implicitly assumes everything is okay and disconnects the avatar after 30 seconds. But if for some reason the viewer never received this (unreliable) message sent via UDP or cannot connect to the new region, this is going to end in a complete disconnect from SL as neither new nor current region are available.

Instead of assuming the TP worked and immediately starting the disconnect process, the current region should only disconnect the current avatar after the new region signals a successful connection with the avatar.

 

[Profaitchikenz Haiku]

10 minutes ago, Whirly Fizzle said:

disconnects the avatar after 30 seconds.

Are you sure it's that much? If I understand the log format correctly, HH:MM:SSZ then 30 milliseconds would be more likely? The lines in the above log file for the failure span 1 second before the stats summary ending.

 

Sorry, my mistake, the disconnect finalises 30 seconds after the two log files diverge, what I was thinking was the divergence is occurring inside the same second.

 

10 minutes ago, Whirly Fizzle said:

Instead of assuming the TP worked and immediately starting the disconnect process, the current region should only disconnect the current avatar after the new region signals a successful connection with the avatar.

Yes please.

Except ... in the event an avatar crashes for some other reason in between the TP start and end (cat chewed through wire) then the sending region isn't going to let go of the avatar and that's where we get the ghosts from?

Edited March 21, 2023 by Profaitchikenz Haiku

[Anna Salyx]

1 hour ago, Profaitchikenz Haiku said:

Yes please.

Except ... in the event an avatar crashes for some other reason in between the TP start and end (cat chewed through wire) then the sending region isn't going to let go of the avatar and that's where we get the ghosts from?

That's mostly handled right now already.  If I'm just hanging out on a region and my viewer crashes for (for any number of reasons including a sudden loss of power or internet), the region is able to detect this fairly quickly* and releases my avatar to wherever orphaned avatars go to.  In a failed teleport were the receiving region never acknowledges the successful arrival, after a reasonable timeout period the original region and simply do a normal crash check routine, and send the remote region a "never mind", so it can clean up anything it needs to.

Now, not knowing the inner workings, what is simple on paper is not always simple in code so it might not be *that* easy, but that's my thought on the matter at least.

----

*for a given value of quickly of course.  I've experienced both crash relogs that let me log in right away and crash relogs that say "yeah, no wait a little bit more" that seems to go on forever.

[Profaitchikenz Haiku]

9 hours ago, Anna Salyx said:

crash relogs that say "yeah, no wait a little bit more" that seems to go on forever.

Now those ones seem to be where the region you were TP-ing from still thinks you are there, and I've hardly seen any of those for months.

[Anna Salyx]

1 hour ago, Profaitchikenz Haiku said:

Now those ones seem to be where the region you were TP-ing from still thinks you are there, and I've hardly seen any of those for months.

No, I was talking about Crash disconnects in general, not TP crashes there.  You know when the viewer crashes to the desktop, or you get a blue screen, or (in my case) my internet stutters (rare but happens), or my router decides it's tired and wants to be rebooted (more common than I care for). All of which just happen when I'm sitting around chatting, not traveling around the grid like a mad woman chasing sales. So I reload my viewer, hit connect, and... sometimes I load right on in.  and..... sometimes it comes back and says that I'm still being logged out try again in a minute.  Which sometimes turns into 5 minutes.....or so

[Anicha Heartsong]

Even on Star Trek, they always had problems with the teleporters - the Heisenberg compensators going off-line and the like.  Perhaps we could build the failed TPs into some elaborate sci-fi roleplaying game... or something.

What worries me is that there might be an alternative virtual reality somewhere where all the avis who crashed on TP end up, wandering around lost or founding evil empires. Spock with a beard anyone?

[Anna Salyx]

 

On 4/3/2023 at 6:18 AM, Anicha Heartsong said:

What worries me is that there might be an alternative virtual reality somewhere where all the avis who crashed on TP end up, wandering around lost or founding evil empires. Spock with a beard anyone?

no no no.  I do not need to add that to my already wild speculations that my avatar is off having a life all on her own while I'm not logged in. You know, traveling the grid and meeting up with other driverless avatars on some exotic sim, sipping piña coladas, and have a generally fully realized life without me. 

And... And!.... and that the main reason I sometimes have difficulty logging in, or the log in take an extra long time, is that my avi is desperately rushing home to be in the last place I logged in, or worse yet, frantically cleaning up the flat like some wayward teenager, and rushing people out the door to hide evidence that a wild party was in progress when I decided to log in. 

No.  I do not need to add anything more like that to my head... alternate dimensions and ghost avatars plotting my demise....

/sighs.  okay too late.

[Anicha Heartsong]

11 hours ago, Anna Salyx said:

my avatar is off having a life all on her own while I'm not logged in. You know, traveling the grid and meeting up with other driverless avatars on some exotic sim, sipping piña coladas, and have a generally fully realized life without me. 

Suddenly it all makes sense. This is why I get all those knowing looks, the strange greetings from men I don't remember meeting. This is why I always seem to run out of Lindens and have an inventory stuffed full of things I don't remember buying. My avi has achieved sentience - and no longer needs me.

I feel bereft.