Bukkit, CraftBukkit, Spigot & BungeeCord 1.13.2

Discussion in 'News and Announcements' started by md_5, Oct 22, 2018.

Thread Status:
Not open for further replies.
  1. gracias sois unos genios!!! :DD
     
  2. MiniDigger

    Supporter

    • Optimistic Optimistic x 1
  3. More from Dinnerbone:

    "As someone asked earlier for git stats, this is the amount of work that happened so far between 1.13 and 1.14's first snapshot (todays): 2717 files changed, 29651 insertions(+), 47402 deletions(-)"

    "and it looks like we need a 'b' already ;) If you close and reopen the F3 overlay you might crash. Don't do that :D "
     
  4. That would mean 1.13.2 is the final update or more to come like 1.13.3? Damn... 1.13.x came like 3 months ago and already 1.14... I'm yet still on 1.12.2... and Forge isn't on 1.13.2 yet... so rip? :/
     
    • Agree Agree x 1
  5. MiniDigger

    Supporter

    this doesn't really mean that its the last 1.13 update. they still could bring out urgent bug fixes if needed. I doubt that will happen tho.

    1.14 is set to release early next year, so there is plenty of time between the two updates. forge is taking so long because they rewrote their entire tool chain and they do major refactorings of the api (and the whole modloader was rewritten too iirc)
     
    • Agree Agree x 2
  6. A bit OT but this means we have to do world conversion once again:
    • Moved light storage from chunks to a separate structure
    • Moved light calculation from all over the code to a self-contained place
    • Moved light computation off the main thread (on the server)
     
  7. [23:54:01 INFO]: Forcing world upgrade! general
    [23:54:01 INFO]: Counting chunks...
    [00:02:25 INFO]: Upgrading structure data...
    [00:02:25 INFO]: 0% completed (0 / 3613818 chunks)...

    When you realise the general/ folder is 3.61 million generated chunks, and even counting it will take nearly ten minutes.

    (4,2 GHz Intel Core i7 / -Xms4G -Xmx16G (of 40GB available ddr4) / NVme SSD)
     
  8. OUCH.... how many GB is the world folder?
    I did a test one on my MacBook (albeit not a fast computer) and 1GB took about 30 min
     
  9. This one world is about 18 gigs, ill get a screenshot in a sec and attach it.
    I'd let it run on the dedicated server, but I noticed I almost get 1,5x the performance on my system.

    The first 10% so far has been 25 minutes, it's at 15% now and has run for 45 minutes.
    [00:46:21 INFO]: 15% completed (542154 / 3613818 chunks)...
    This might give a tiny indication as to what to expect.

    Smaller worlds and temporary worlds are faster to just create new, like a /spawn world on almost all our servers, or the /test worlds and /mine worlds. We can just generate those fresh. And the /spawn worlds are so small it's not a big deal to convert those. It's done in 15 minutes tops. Any temporary world we at some point will just unload and not introduce 'right away' back into the converted setup so we can have very little downtime. And run those on a clone and then later re-import those as a new world and update the plugins to point to it again etc.
    Bigger worlds like this one that are 12gb or more and we never remove, they've been with us from 2011 and upwards, we give those priority with testing.

    I am actually considering to leave them as they are but tag them as /legacy worlds. So players still have full access, but then have fresh new 1.13.2 and up worlds as the new general / wilderness / towny etc. And keep it small - push new players to those, and allow all players to use them. I haven't really decided yet on that.

    Screenshot 2018-10-25 at 00.45.20.png

    My biggest fear is that things all convert, we use it - things seem fine. Then 3 or 6 or 9 weeks later the whole setup starts crashing because a world has corrupt chunk due to nbt data or whatever and I find out I have no idea how to properly fix that in 1.13.2.

    There's a creative world that's 75x without pre-rendered 1.12.2 chunks, from old 1.8 setup, that map is already 71 gb. It's 6 years old, and had issues in the past with updating to 1.10, etc. i am not looking forward to even test that one.

    And since I am a noob, and just doing my best here, I am not a developer either, so just throwing sticks at the screen when stuff doesn't work. It's actually quite stressful. I have some experience, and this isn't the first time we deal with bigger updates like this. But it's been a while since it's been this stressful. I also don't have money lying around to just hire a box somewhere with 128gb ram and dual xeon cpu's etc for a few months to force update everything to the max. We do this with 2 ppl, the support from this community, and hopefully some common sense and learning from the mistakes, just using the resources that we have.

    [edit]
    [01:00:00 INFO]: 19% completed (714973 / 3613818 chunks)...
    [01:00:01 INFO]: 19% completed (715175 / 3613818 chunks)...
    [01:00:02 INFO]: 19% completed (715349 / 3613818 chunks)...
    10% for the first 25 minutes, almost 20% for the full hour.
    This seems to be at least 5 hours .. we will see how the performance is after a few hours - has it dragged to almost a stop, or will it hold up?

    [edit]
    ```[02:28:04 INFO]: 50% completed (1830841 / 3613818 chunks)...```
    2,5 hours for 50% we're on schedule.. seems like we're still on for 5 hours for 100%

    it's not showing a sign of slowing down, it's the same as the first 10% .. and 3,6m chunks for this hardware setup seems to be 5 hours for a 18gb map.
     
    #91 mrfloris, Oct 24, 2018
    Last edited: Oct 25, 2018
  10. I've started updating plugins with the hope of moving from 1.8.8 to 1.13.2 at some point. It's so stressful, it's making me physically ill. Which is certainly not worth it considering only a handful of kids play on the server, only one of which I know IRL.

    The plugin causing me the most stress and the most wasted time (hours and hours and hours) is PlotSquared. What a piece of crap.
     
    • Agree Agree x 2
  11. One thing I've done is make sure any old 1.8 plugin that (kinda) works on 1.12.2 is updated to a 100% working on 1.12.2 but that has been reported to work with 1.13.x as well, or is in the making. Making sure that updating those plugins going from 12 to 13 will go a lot smoother.
    Any plugin that has poor 1.13 support, is too old or already had issues with going from 1.8 to 1.12, we're looking to find modern alternatives for them. And then i look for 1.12.2 plugins that have 1.13.1(2) support. And test updating what we have on 1.12.2 first. i won't upgrade until what we run on 1.12.2 has 1.13.2 alternatives (and has been tested etc).

    But it's a stressful thing indeed. I take it as it comes. And we run a stable 1.12.2 setup, nothing is 1.8 anymore. This year really was getting rid of 1.8 stuff that kinda runs on 1.12.2, and make sure it at least runs on 1.12.2 and has an active dev etc. It's the first time I am also considering not finding a 'solution' to get groupmanager to work well, but rather see if we can properly convert to LuckPerms, since it is active, widely used, seems to get good support, and has 1.13.x support as well. So the update from 1.12.2-groupmanager->luckperm, and then 1.12.2 luckperm to 1.13.2 luckperm should go smooth. Another thing we're doing is seeing if there are plugins we're using that offer a feature, but that things like CMI might have as a feature. So we can finally get rid of say 5+ plugins, and let CMI do it. Since cmi works fine on 1.13.x it is nice to make that shift now in 1.12.2 and let users get used to it, then when we move to 1.13, it's 5+ plugins less that i have to test and worry about.
     
    • Agree Agree x 1
  12. My initial approach was to jump from 1.8.8 to 1.13 with all updated/replaced plugins. I never got PlotSquared to work (we use PlotMe on 1.8.8).

    So now I'm trying to update/replace the plugins under 1.8.8 with the hope that jumping to 1.13.2 will be easier:
    • Essentials -> EssentialsX
    • GroupManager -> LuckPerms
    • Prism -> CoreProtect
    • Updated SilkSpawners (and updated my spawner shop buy signs)
    • Updated Vault
    • Updated MobArena
    Next is to try PlotMe -> PlotSquared. I expect PlotSquared to crash and burn during the conversion. I'm probably going to have to write a program to go through the PlotMe database and create a script I can use to issue claim and trust commands for 500 plots.
     
  13. btw, 5 hours later, ..

    [04:49:56 INFO]: 99% completed (3613649 / 3613818 chunks)...
    [04:49:57 INFO]: World optimizaton finished after 17252345 ms

    I think it's spelled: optimization
     
    • Like Like x 1
    • Winner Winner x 1
    • Friendly Friendly x 1
  14. It was a ton of work -> optimizaton :ROFLMAO:
     
    • Creative Creative x 1
  15. Sounds like your bottleneck is computer horsepower and the fact you can only process one world at a time. Why not find other community members who can spare the CPU cycles, ZIP up a world or two, and have that person convert the world then ZIP it and send it back?
     
  16. md_5

    Administrator Developer

    Its done already though :p
    Would take longer to try and coordinate that
     
    • Agree Agree x 1
  17. I don't run plugins for this very reason.
     
    • Optimistic Optimistic x 1
  18. MiniDigger

    Supporter

    for
    99% of servers thats not an acceptable solution to that problem, lol
    as long as you use active plugins and don't make giant jumps like from 1.8 to 1.13 its just fine to run plugins.
     
Thread Status:
Not open for further replies.