Cool VL Viewer forum

View unanswered posts | View active topics It is currently 2024-03-28 18:00:52



Reply to topic  [ 20 posts ]  Go to page 1, 2  Next
Mesh objects fail to rez with object cache active 
Author Message

Joined: 2011-12-13 14:11:38
Posts: 186
Reply with quote
I've got quite a few objects failing to rez at my home and it seems to be related to whether the object cache is active or not: when it is, some objects fail to rez, I can't see them at all. If I turn the object cache off and relog, everything appears correctly. But if I turn the cache back on and relog, the objects fail to rez again. It actually seems to impact most mesh objects apart from those attached to my av: even the buildings outside fail to rez. I've attached the logs of a session with object cache turned on, and another one with object cache turned off; in both sessions, I just logged in and waited for things around me to appear.
I tried to tweak various options related to the object cache, but I couldn't find anything that had an impact. I also tried to remove the folder named "objectcache" both in .secondlife/cache & .secondlife/cache_coolvlviewer, but it hardly changes anything: objects still don't appear after a relog if object cache is turned on. I also tried to force everything to appear via Advanced -> Rendering -> Refresh visibility of object, but that didn't do anything.
Is this an issue with Cool VL Viewer, or is there something wrong with one of my settings or something?
Thanks!


Attachments:
logs.zip [53.52 KiB]
Downloaded 89 times
2021-12-28 20:14:26
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
This is not a viewer bug. It affects all viewers recently...

See the corresponding topics on the SL forums. For example:
https://community.secondlife.com/forums ... -of-there/
https://community.secondlife.com/forums ... lickering/
https://community.secondlife.com/forums ... verything/

There is probably something fishy with the server-side "interest list" code (perhaps changes related with LL's 360° viewer ?)...

By disabling the object cache (which code did not change in months), you simply ignore the interest list... which "cures" the problem.

Perhaps you could also try enabling the 360° interest list option in the Advanced -> Cache menu, and see if it makes any difference.


2021-12-29 00:31:18
Profile WWW

Joined: 2011-12-13 14:11:38
Posts: 186
Reply with quote
Ah OK... Well, I did try with Firestorm and it looked like it was working, but since I never use it, I have no idea what options are on by default, so maybe I just got lucky.
And I did try to enable the 360° interest list, but it didn't make any difference.
I'll monitor the topics you mentioned to see if/when the issue gets fixed. Sorry for the noise, and thanks for the answer! :)


2021-12-29 18:53:07
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
ErikaThorkveld wrote:
so maybe I just got lucky.
The occurrence is totally random. Some people appear to encounter it more often than others (perhaps because of a specific sim server version ?). I personally only encountered it once (while riding a mesh vehicle that suddenly "disappeared" even though my avatar was still riding it).

EDIT: in the event you could consistently reproduce the issue with a given mesh, I'd be curious to see it myself (and try to diagnose it doing so)...


2021-12-29 19:31:19
Profile WWW

Joined: 2011-12-13 14:11:38
Posts: 186
Reply with quote
Henri Beauchamp wrote:
in the event you could consistently reproduce the issue with a given mesh, I'd be curious to see it myself (and try to diagnose it doing so)...

If it works for you like it does for me, reproducing the issue should be fairly easy: I just took 2 photos of how things appear when I'm at this location:
https://maps.secondlife.com/secondlife/ ... 5/178/2501
This is the one with object cache off, so this is how it should appear:
Attachment:
QuantumCity_001.jpg
QuantumCity_001.jpg [ 57.89 KiB | Viewed 1829 times ]

This is the one with object cache turned on:
Attachment:
QuantumCity_002.jpg
QuantumCity_002.jpg [ 31.03 KiB | Viewed 1829 times ]

Here are the options for the object cache in the Advanced -> Caches menu:
  • Object cache: on
  • Full region caching: on
  • Use object cache occlusion: off
  • Bias objects retention: off
  • Balance object cahce: on
  • Force 360 degrees interest list: off

By the way, just after I turned object cache on, I got a crash when trying to relog - this had sometimes happened already when I was testing the object cache. This happens only the first time I try to relog, if I try again, the crash no more happens. I'm attaching the log & the stack trace.


Attachments:
crashinfo.zip [19.14 KiB]
Downloaded 88 times
2021-12-30 06:08:15
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
ErikaThorkveld wrote:
If it works for you like it does for me, reproducing the issue should be fairly easy: I just took 2 photos of how things appear when I'm at this location:
https://maps.secondlife.com/secondlife/ ... 5/178/2501
It appears just fine for me: I tried with or without object cache, after a TP to or a relog in that same location, with WL and EE renderers, ALM and shadows on or off (all combinations)... Here is what I get (in this case in WL mode, no ALM):
Attachment:
Citrine.jpg
Citrine.jpg [ 68.92 KiB | Viewed 1826 times ]
And this is with EE + ALM + shadows (after a relog):
Attachment:
Citrine2.jpg
Citrine2.jpg [ 73.14 KiB | Viewed 1823 times ]


Quote:
Here are the options for the object cache in the Advanced -> Caches menu:
  • Object cache: on
  • Full region caching: on
  • Use object cache occlusion: off
  • Bias objects retention: off
  • Balance object cahce: on
  • Force 360 degrees interest list: off

"Use object cache occlusion" should be on, "Bias objects retention" is best kept on as well (these are the default settings), however even with them off, I still do not see the slightest issue...

Quote:
By the way, just after I turned object cache on, I got a crash when trying to relog - this had sometimes happened already when I was testing the object cache. This happens only the first time I try to relog, if I try again, the crash no more happens.
This is interesting... I do not get any crash, and the crash does not seem directly related to the object cache either as it happens in the texture cache instead... I never encountered such a crash either and no one else ever reported one so far... To help me diagnose the crash, could you please enable the "APR" debug tag via the "Advanced" -> "Debug tags" of the login screen menu, and should you crash, post the log here.

To try and solve the rezzing issue, I recommend to:
  • Close the viewer.
  • Rename or move your ~/.secondlife/user_settings/settings_coolvlviewer_1282.xml file.
  • Relog, observe and report.

Also, did you try the "Refresh visibility of objects" feature (in "Advanced" -> "Rendering"; shortcut is ALT SHIFT R) ?... If it makes the objects reappear, then try increasing the value for the "ObjectsVisibilityAutoRefreshDelay" debug setting to 3, 4 or 5 seconds (should be more than enough, even if the default 2 seconds were too short): this governs how long after a login, a TP, or a region border crossing, the refreshing feature gets automatically triggered.

Finally, have you verified that your disk (where the cache files are stored) is error-free ('smartctl -a /dev/sdX' for info, 'smartctl -t short /dev/sdX' for a quick test, with 'smartctl -a /dev/sdX' again after the delay requested for the testing to finish) ?

EDIT: I also see that you are using an old Linux kernel version (5.3), which is not even a long term branch and thus no more maintained; this should be fine nonetheless, but one never knows, and you might want to update, to at least v5.4 (closest long term branch), and preferably v5.10 (would bring io_uring support, which the Cool VL Viewer can use thanks to Kathrine Jansma's contribution).


2021-12-30 09:58:46
Profile WWW

Joined: 2011-12-13 14:11:38
Posts: 186
Reply with quote
Damn it! I did try everything you said and there wasn't any change, but after I renamed my settings file and relogged, in addition to the objects not rezzing, I saw that a lot of textures appeared to be corrupted. So I did the one thing everybody does when this kind of things happen: I cleared my cache and relogged. And lo & behold: with object cache on, everything's fine now... I'm reading all the time that clearing one's cache is rarely the proper solution, but this time, it seems it was. :lol: I must have had something fishy in it that caused the issue, and probably the crash I got as well. If I ever get the crash again, I'll try to reproduce it with APR on in the debug tags.

And by the way, I did test my disk and it looked fine (got messages like "SMART overall-health self-assessment test result: PASSED" & "No Errors Logged", so even if I'm far from understanding everything it says, I'd say it looks OK :D ). Not sure why/how the cache got corrupted...

Well, sorry again for the noise. It looks like it wasn't the problem everybody's getting after all...

Anyway, thanks a lot for your help! Not sure I would have figured it out without your advice. :)


2021-12-30 19:50:57
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
ErikaThorkveld wrote:
Damn it! I did try everything you said and there wasn't any change, but after I renamed my settings file and relogged, in addition to the objects not rezzing, I saw that a lot of textures appeared to be corrupted. So I did the one thing everybody does when this kind of things happen: I cleared my cache and relogged. And lo & behold: with object cache on, everything's fine now... I'm reading all the time that clearing one's cache is rarely the proper solution, but this time, it seems it was. :lol: I must have had something fishy in it that caused the issue, and probably the crash I got as well.
The caches (assets cache excepted) got index files (files that contain headers and/or sundry data for the actual data "bodies" that are held in separate files): when a crash occurs, the cache index may get out of sync with the cached file if the crash happens while some data was not fresh in it and being updated. In this case, only a full cache clear can solve the corruptions seen on next relogs...

Quote:
If I ever get the crash again, I'll try to reproduce it with APR on in the debug tags.
Thanks. The "TextureCache" debug tag could also be enabled, for even more infos. The crash was likely due to the data corruption, but it is not 100% sure either and this could be the result of a weird race condition...

Quote:
And by the way, I did test my disk and it looked fine (got messages like "SMART overall-health self-assessment test result: PASSED" & "No Errors Logged", so even if I'm far from understanding everything it says, I'd say it looks OK :D ).
Things to watch for in the 'smartctl -a' data are the entries such as "Reallocated_Sector_Ct" (HD), "Runtime_Bad_Block" (SSD), "Uncorrectable_Error_Cnt", "Current_Pending_Sector", "Offline_Uncorrectable", "Multi_Zone_Error_Rate" (the actual names and the presence of these stats may vary depending on the type of drive and/or model): if these numbers "RAW_VALUE" are anything else than 0, then it may reflect a dying drive... The "PASSED" info is usually unreliable (for example HD drives makers might consider that a moderate number of reallocated/pending sectors are OK and "passed", while they do imply that the drive is unreliable; such a drive will likely get automatically ejected from a RAID array, for example, and such sector failures will cause data corruption anyway).


2021-12-30 20:06:04
Profile WWW

Joined: 2011-12-13 14:11:38
Posts: 186
Reply with quote
New development: everything was working fine with object cache on, until earlier today, when I got disconnected from SL. Apparently, there was a problem with the internet connection, so everything froze in-world; I noticed my connection was gone, and I quit SL. And when I logged back in, the problem was back: a lot of objects around me would not rez, and disabling the object cache made them reappear. So if the problem was indeed caused by a cache corruption, could it be possible that the disconnection caused it? This kind of disconnection happens regularly to me, so it might have been the cause of the earlier cache corruption too.


2022-01-01 07:57:33
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
ErikaThorkveld wrote:
could it be possible that the disconnection caused it?
Yes, it is likely an explanation for what you experienced. I will have a look at the LLVOCache code, to see what can be done to avoid such issues (perhaps by disabling the cache writes on such disconnections).

Quote:
This kind of disconnection happens regularly to me, so it might have been the cause of the earlier cache corruption too.
It happens to me as well (and to everybody), but I so far never experienced an object cache corruption in such cases. Perhaps a race condition that happens for you and not for me...


2022-01-01 09:07:29
Profile WWW
Display posts from previous:  Sort by  
Reply to topic   [ 20 posts ]  Go to page 1, 2  Next

Who is online

Users browsing this forum: No registered users and 15 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group
Designed by ST Software.