#musicbrainz-devel

nikki: it doesn't touch the database

oh?

I was going to complete ignore the DB and remove the column at the next schema change

my branch simple loads recording->length from track lengths, and never edits it

we haven't decided to get rid of standalone recordings, so we can't do that :/

simply*

oh

this is on a PC, Fusion seems to be Mac only

guess I'll need to rework it then

ruaok: bin/replicate now finished

ijabz: on windows you want "VMWare Player"

warp, ah thx doesnt even seem to be listed on mian product page but found it now

ianmcorvidae: tell me if you get it to work with VirtualBox. I didn't and I am out of ideas currently. Also added a comment to the blog post.

yeah, I'm poking at it

but who knows what, if anything, I'll figure out

out of interest does Vm Fusion just turn your mac into a linux/windows box stopping its usage as a OSX box ?

no

at least, none of the ones I've used have done that

ijabz: if I remember correctly, the only special thing about fusion as opposed to normal vmware is that it let you have individual windows per app rather than a desktop window that hold everything.

ruaok: so far bin/reindex seems to be working. I'm going to suspend it now, I'll resume it in a couple of hours and leave it running then.

oh,ok

great, thanks!

kepstin-work: yeah, sounds about right

I just run it as a normal single window though

um, username:musicbrainz, password:musicbrainz doesnt seem to be correct ?

username: vm, password: musicbrainz

ah I was looking at doc page rather than wiki page

oh is the transcluded page out-of-date? nikki, reosarevok, can either of you check/transclude?

heh for me | command ouputs >, does everyone get that, how can I fix that

the '|' key gets you a '>' character you mean?

what page? I don't see anything about a vm in the transclusion table that's out of date

nikki: it'd be server setup, I think

nikki: http://musicbrainz.org/doc/MusicBrainz_Server/S...

nikki: I don't think we have a dedicated VM page or whatever

updated

ijabz: the virtual machine is probably configured for a US Qwerty keyboard layout.

ok, ill look later

much easier to just SSH in

then you'll use your natural keyboard layout, not the machines

well good news for me server starts up and I can connect to it okay

that is good news

:)

virtualbox seems to be failing to even issue the same error message repeatedly, for me :P

Ive given the VM 8GB, and just kicked of the search index building

ruaok: we should update the blog post to say this doesn't work in VirtualBox

nikki: I am not really sure the documentation should be updated already, since the new VW at least doesn't work with VirtualBox

ijabz: you rebooted it too, right?

ijabz: could you run: psql -U postgres template1 -c "SHOW shared_buffers"

(first time, importing just the vmdk, it says something about a misformatted header, but after that it seems to just say it's unsupported)

I set it to 8GB before i started it

wanna make sure the configuration thing worked

ijabz: ok, cool

JonnyJD_: the old vm is pretty useless anyway

nikki: hm, because of the perl dependency hell thing?

it's probably only useful if you actually deliberately want out-of-date data

well, because it's ancient

ocharles 1995MB

ijabz: perfect

so thats right I guess

ruaok's script is doing the right thing :)

seems to be working okay, though I'm a bit disappointed with this labtops sound and screen, performance wise no problems

so, aside from VB issues, no problems so far?

not for me (unles syou include the key mapping)

"One of the things I consider before diving into a new library is whether or not ocharles has written an intro for it." awwwwww

the primary difference I can see in the vmdk headers is that the new one seems to have spaces after values, where an old vmdk from the last release doesn't

by the by

re: the "improperly quoted value" thing, I think, but

other than some other random differences, like ours claims to be readonly, it has an embedded name thing of "generated-stream.vmdk" rather than the real name, uses some different adapter types -- but none of that has to do with actual header formatting

post updated with vb caveat

bye JIRA

yeah, thats webalizers's fault.

wat

webalizer breaks jira?

reading the logs freezes the machine?

it pegs the CPU

seems to be the issue, anyway

we have huge logs and a complicated webalizer config

the two interact poorly :P

heh, oops, full disk

heh, appears to have worked

I mean, and created a 27GB file, but :P

ianmcorvidae: no dynamic disc anymore (so the disc is really that big anyways?)

well, it's a 40GB disk

so 27GB is at least... better than that

this appears to be working, anyway

VBoxManage clonehd <the extracted vmdk> something.vdi --format VDI

and then you can import the vdi

possibly there's some way to make the VDI less space-intensive though

seems to be about 18G of disk actually used, anyway

turn the disk into a raw file image, load it into qemu on linux, use 'fstrim' to free the space, then convert it back :)

haha

VBoxManage modifyhd seems to have a --compact option, which seems likely to be what's needed

yea, but that doesn't work with VMWare

my original script already called that

the --compact option can't help with areas on the fs that contain data that's not referenced, I think?

(when building the separate DB drive)

well, this is using a vdi anyway, since the vmdk that VMWare output fails entirely on vbox

kepstin-work: not sure about that, probably not

I'm mostly just exploring options right now though

"For this operation to be effective, it is required that free space in the guest system first be zeroed out using a suitable software tool." - so, yeah.

but using the raw image format with qemu on linux is nice, because it supports the TRIM operation, which turns into holes in the sparsely allocated file.

assuming a new enough linux kernel to have the 'hole punch' operation.

ianmcorvidae: you managed to use VBoxManage on the same vmdk that you downloaded?

I get VBoxManage: error: VMDK: incorrectly quoted value in descriptor when running VBoxManage clonehd

I did, yes

well, I mean, I copied it to a .bak first and then cloned that

ruaok, ijabz: "recording: Unable to get replication information" when running bin/reindex, something to worry about?

nope.

ianmcorvidae: maybe you did something to that vmdk (without knowing). I can't get clonehd to work.

possibly

my virtualbox did do some editing of the vmdk, but not the one I was using

(as in, I blew everything away, did a fresh untar of the ova, then copied to a .bak, which is what I was clonehd'ing)

f27ec2db10aef6cabfc62d3209907f4c MusicBrainz_2013-08-01-disk1.vmdk

hm, mine is definitely different

8c994faf3075bc7ffc8fd34db19fb446 MusicBrainz_2013-08-01-disk1.vmdk.bak

JonnyJD_: could you hexdump -C MusicBrainz_2013-08-01-disk1.vmdk | head -n 50 and pastebin it?

ianmcorvidae: http://pastebin.com/VneEEnub

that's not the same image that's in the ova, I don't think

because it has the ddb.uuid.modification stuff, etc.

https://pastee.org/jprzs is the header of the one that worked successfully for me

https://pastee.org/wmpct is the one that failed to import in virtualbox

(which as you can see is similar to yours, just with different UUIDs)

ianmcorvidae: I'll try and extract from the ova again

Yippie, build fixed!

Parameters:

ghprbActualCommit=c0413819b77883798e3f30b5c4f578c94b25ab73

ghprbTargetBranch=master