> -----Original Message----- > From: Christopher Schultz [mailto:ch...@christopherschultz.net] > Sent: Tuesday, August 06, 2013 10:52 PM > To: Tomcat Users List > Subject: Re: [OT] Using the bin/daemon.sh script on ubuntu. > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA256 > > Igor, > > On 8/6/13 6:28 AM, Igor Cicimov wrote: > > On 06/08/2013 12:40 AM, "Jeffrey Janner" > > <jeffrey.jan...@polydyne.com> wrote: > >> > >>> -----Original Message----- From: Christopher Schultz > >>> [mailto:ch...@christopherschultz.net] Sent: Friday, August 02, > >>> 2013 10:30 PM To: Tomcat Users List Subject: Re: [OT] Using the > >>> bin/daemon.sh script on ubuntu. > >>> > >>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 > >>> > >>> Christian > >>> > >>> On 8/1/13 11:55 AM, Christian Schneider wrote: > >>>> On our (AWS) installation we have limited space on /opt, therefore > >>>> we attached an EBS volume to /var/, - otherwise we would get > >>>> problems with the log files. Now it can grow above some GB. > >>> > >>> Have you thought about using /mnt/ephemeral[0-9]? > >>> > >>> Our instances have ~1TiB in combined ephemeral storage available > per > >>> instance. I'm sure it depends upon the instance type, though. > >>> Remember that terminating the instance loses the ephemeral data.. > >>> that's what makes it ... ephemeral. Just remember to copy what you > >>> need back to an EBS-backed storage volume before you terminate. > >>> > >>> - -chris > >> > >> Chris, If I remember my empirical testing of the AWS ephemeral > >> storage system, you actually lose data on shutdown, not termination. > > > > Not true, the ephemeral data is only lost on instance termination. > > Confirmed -- at least with "shutdown -r", and on an m1.large instance > - -- /mnt/ephemeralX retains files across reboots. I didn't do a "stop" > and "start" from the GUI or anything like that, though. > > I think you're supposed to be able to rely on ephemeral storage. The > thing is that it's not EBS-backed: it's actually only on the physical > server that you get when you launch an instance. When you shut-down, > the data stays on the physical machine and you can only get to it again > when you boot the same instance on the same machine. Since AWS always > runs instances on the same physical hardware every time, you're good > unless you /terminate/ (i.e. discard) the instance. > > Honestly, I was a bit surprised to find out that each time you launch > an instance it doesn't just get launched on some random piece of > equipment. That would seem more flexible (if wasteful: moving GiBs of > data around to various hosts isn't exactly fast) to me. > > On more than one occasion, I've had an instance start acting funny and > shortly thereafter, I get an email from AWS saying that they will be > terminating my instance in a few days. You can't just "migrate" the VM > to another piece of hardware. Instead, you have to image the VM, create > an AMI from that image, create a new instance with that AMI, and > destroy the old instance. (Then delete the old image *and* EBS store, > otherwise you continue to pay for the disk space taken up by an > instance that will never again be launched). Pain in the neck. > > I think you can create an AMI directly from a running instance these > days (probably for this exact reason). > > - -chris
Well, I must admit my test was using an m4.xlarge with EBS-backed root, so my tests might have had some difference. I don't remember if I tested it with "shutdown -r" or "init 6", though I do remember doing a stop/start from the GUI (since that's probably what my backup folks here would end up doing). BTW: If you set up your instance with an EBS-backed startup drive (or all useful drives), then you don't have to worry about all that AMI business, they can just mount your drives on another physical box. Of course, at that point, the ephemeral drives are basically useless for anything other than temp/swap space. Makes one wonder why they are even offered, though I'm OK with using them for those purposes. Jeff
