Amazon EC2 t1.micro can't run HHVM

[makersunion]

Server config:
EC2 t1.micro Ubuntu12.04 64bit

HHVM version:
HipHop VM v2.1.0-dev (rel)

Error alert is :
tcmalloc: large alloc 1209565184 bytes == (nil) @
tcmalloc: large alloc 1209565184 bytes == (nil) @
could not allocate 1209565183 bytes for translation cache

And I've tried this: b5163ad
But, still not work.
Help me, please.

[scannell]

What happens when you try that? Are you sure you specified those settings correctly? (It should be complaining about a different number of bytes at least.)

[makersunion]

No, always show :could not allocate 1209565183 bytes for translation cache .
And I used code :
Eval {
JitASize = 256 << 20 # 256MB
JitAStubsSize = 256 << 20 # 256MB
JitGlobalDataSize = JitASize >> 2 # 64MB
}
When I try : hhvm -m daemon -u ubuntu -c hhvm.hdf show this alert.

[scannell]

Try passing the actual numbers? I don't think configuration file arguments are evaluated...

[makersunion]

Actual numbers mean is JitASize = 256M ?

[scannell]

An integer number of bytes, e.g. 268435456.

[makersunion]

There is no change. Eval in Server{ }? or Server { } Eval { }?

[scannell]

Should just be under Eval { }.

[makersunion]

No work :(

[scannell]

Try passing the arguments on the commandline, or just attaching a debugger in the runtime option parsing code and seeing if it's finding the values -- this should narrow down the problem.

[makersunion]

Eval {
JitASize = 268435456
JitAStubsSize = 268435456
JitGlobalDataSize = 67108864
}
Eval.Debugger {
EnableDebugger = true
EnableDebuggerServer = true
Port = 80
DefaultSandboxPath = /var/www/
}

Just have this , no change.

[makersunion]

hhvm --debug-config hhvm.hdf
No change. :(

[scannell]

I verified that using this in a .hdf file:

Eval {
JitASize = 1024
JitAStubsSize = 1024
JitGlobalDataSize = 100
}

That the VM terminates because the sizes are too small, so these values are taking effect. Are you sure that .hdf file exists? I'd try stepping through the code.

[makersunion]

Yeah,
root@ubuntu:/etc# hhvm -m daemon -u ubuntu -c hhvm.hdf
tcmalloc: large alloc 1209565184 bytes == (nil) @
tcmalloc: large alloc 1209565184 bytes == (nil) @
could not allocate 1209565183 bytes for translation cache
I'm sure /etc/hhvm.hdf is OK

[makersunion]

This is same question: http://www.swageroo.com/wordpress/hiphop-vm-on-ec2-could-not-allocate-1210089471-bytes-for-translation-cache/ m1.small is OK ,but t1.micro not work. And Amazon EC2 t1.micro memory is 678MB

[scannell]

What happens when you step through the code?

[makersunion]

Yeap~!
When I used this code :
Eval {
JitASize = 1024
JitAStubsSize = 1024
JitGlobalDataSize = 100
}
alert:
Log file not specified under daemon mode.\n\n

[makersunion]

Look like is worked.Wait ... I try it now.

[makersunion]

uh... :(
root@ubuntu:/etc# hhvm -m daemon -u ubuntu -c hhvm_debug.hdf
Log file not specified under daemon mode.\n\n
root@ubuntu:/etc# hhvm -m server
mapping self...
mapping self took 0'00" (54830 us) wall time
loading static content...
searching all files under source root...
analyzing 728 files under source root...
loaded 0 bytes of static content in total
loading static content took 0'00" (22046 us) wall time
tcmalloc: large alloc 1209565184 bytes == (nil) @
tcmalloc: large alloc 1209565184 bytes == (nil) @
could not allocate 1209565183 bytes for translation cache

No work...

[scannell]

Of course the second time didn't work, you didn't specify an .hdf file...

[makersunion]

How to specify an .hdf file?

[makersunion]

hhvm -m server -u ubuntu -c hhvm_debug.hdf ?

[makersunion]

Allocation sizes ASize, AStubsSize, and GlobalDataSize are too small.

[makersunion]

I try 512MB to byte?

[scannell]

If you got the 'too small' portion, you specified it correctly. Now change the values to sizes that will fit on your VM.

[makersunion]

I try config for t1.micro use this code:
Eval {
JitASize = 1149304830
JitAStubsSize = 1149304830
JitGlobalDataSize = 608206847
}
But ,alert this:
Combined size of ASize, AStubSize, and GlobalDataSize must be < 2GiB to support 32-bit relative addresses
What does this mean?

[makersunion]

Yeap~！ It's worked
t1.micro use this code:
Eval {
JitASize = 134217728
JitAStubsSize = 134217728
JitGlobalDataSize = 67108864
}

I want to try the most appropriate distribution ratio for Amazon EC2 t1.micro.
Thanks scannell very much.
I am very grateful for that.

[scannell]

Glad you got it working.

[kstan79]

Can I know compiling hhvm under Amazon ec2 micro take how long time.

[karpa13a]

@kstan79 use ubuntu packages
@sgolemon maybe it's time to make "nightly" builds repo?

[scannell]

In general we've been talking about other release/packing strategies but haven't made any concrete decisions yet. (In the meantime, anyone is free to set up nightly builds for us from the available source if they would like.)

[karpa13a]

@scannell can you submit in master or in this issue debian/ directory contents for self build

[scannell]

@sgolemon is the right person to ask for that -- I'm not entirely sure how she packages these releases.

[jrborbars]

I'm trying to run hhvm in the t1.micro instance (AWS). The messsage appear when I tried to run:

sudo /usr/bin/hhvm --mode daemon --user web --config /etc/hhvm.hdf

from http://www.hhvm.com/blog/113/getting-wordpress-running-on-hhvm.

alert:
Log file not specified under daemon mode.\n\n

This is related with:

Eval {
JitASize = 134217728
JitAStubsSize = 134217728
JitGlobalDataSize = 67108864
}

Where I put this code? In server.hdf or config.hdf?

Thanks a lot.

[scannell]

I'm not sure if you still need to change the JIT sizes after 5d60ea9. The alert is telling you just that -- I'm guessing /etc/hhvm.hdf doesn't specify a log. The name of the .hdf file doesn't matter, the only thing that matters is if you want HHVM to use it you need to specify it as the --config argument when starting HHVM.

[jrborbars]

Ok. But /etc/hhvm.hdf doesn't exists anymore. When I install from binaries (sudo apt-get install hhvm) three files are created under /etc/hhvm directory: config.hdf, server.hdf and php.ini (this later is an empty file). My instance seems running (after run sudo /usr/bin/hhvm --mode daemon --config /etc/hhvm/server.hdf), but I pointed to my index.php under app directory and the only response is: not found.

What I miss?
Thanks for your quick response too.

P.S.: the header of response is (from firebug):
Content-Length 9
Content-Type text/html; charset=utf-8
Date Fri, 03 Jan 2014 23:21:50 GMT
X-Powered-By HPHP

[scannell]

I believe in that article it says to create /etc/hhvm.hdf using the sample config file in the article. You can either use the inbuilt ones that ship with the binaries or the one in the article -- pick one and use that as the --config argument.

[jrborbars]

Yes, I picked one (server.hdf, provided as is). In one hand the server appear to be running, but in the other hand none of the php code was processed, and I don't know why.

[scannell]

For wordpress i would use a config file with the contents of the one on the wiki, otherwise you probably need to make sure all the paths are correct.