Re: [systemd-devel] hinting systemd about what to do to make a device appear

also sprach Lennart Poettering [2015-09-11 14:16 +0200]: > I figure the next step is to attach gdb to PID, set a break point > in device_process_new() and figure out why systemd doesn't care > for the symlink. Frankly — and you won't like this — but I lack the resources

Re: [systemd-devel] hinting systemd about what to do to make a device appear

Отправлено с iPhone > 10 сент. 2015 г., в 20:21, martin f krafft написал(а): > > also sprach Michael Biebl [2015-09-10 19:13 +0200]: >> Could you run udevadm info on the device? >> Does it have a SYSTEMD_READY flag set? > > No:

Re: [systemd-devel] hinting systemd about what to do to make a device appear

also sprach Andrei Borzenkov [2015-09-11 10:28 +0200]: > Please run "udevadm monitor --env", start drbd service to trigger > device creation and make output available. http://slexy.org/raw/s21alRk3ZD -- @martinkrafft | http://madduck.net/ | http://two.sentenc.es/ "if you

Re: [systemd-devel] hinting systemd about what to do to make a device appear

also sprach Lennart Poettering [2015-09-11 12:30 +0200]: > Do I see this right, this is luks on lvm on drbd? Yuck! luks on drbd, and the local volume is on lvm, yes. > Which symlink are you looking for? the drbd one? Right. /dev/drbd/by-res/backup -- @martinkrafft |

Re: [systemd-devel] hinting systemd about what to do to make a device appear

On Fri, 11.09.15 12:45, martin f krafft (madd...@madduck.net) wrote: > also sprach Lennart Poettering [2015-09-11 12:30 > +0200]: > > Do I see this right, this is luks on lvm on drbd? Yuck! > > luks on drbd, and the local volume is on lvm, yes. > > > Which symlink are

Re: [systemd-devel] hinting systemd about what to do to make a device appear

On Fri, 11.09.15 10:41, martin f krafft (madd...@madduck.net) wrote: > also sprach Andrei Borzenkov [2015-09-11 10:28 +0200]: > > Please run "udevadm monitor --env", start drbd service to trigger > > device creation and make output available. > >

Re: [systemd-devel] hinting systemd about what to do to make a device appear

On Thu, 10.09.15 12:00, martin f krafft (madd...@madduck.net) wrote: > Hi, > > on one of my machines, /srv/backup is an encrypted filesystem > sitting on top of a DRBD volume. In /etc/crypttab, I thus have > > backup /dev/drbd/by-res/backup none noauto > > and then I mount /dev/mapper/backup

Re: [systemd-devel] hinting systemd about what to do to make a device appear

On Thu, Sep 10, 2015 at 12:14:56PM +0200, martin f krafft wrote: > also sprach Richard Maw [2015-09-10 12:13 > +0200]: > > I've seen this happen when udev is misbehaving because your kernel doesn't > > have > > CONFIG_FHANDLE enabled. Can you check that first? > >

Re: [systemd-devel] hinting systemd about what to do to make a device appear

also sprach Lennart Poettering [2015-09-10 12:28 +0200]: > Does udev know about the symlink? I.e. does "udevadm info" show it? > Which component is creating it? Only symlinks created by udev are > known by udev and systemd. To support this, the drbd support needs to >

Re: [systemd-devel] hinting systemd about what to do to make a device appear

also sprach Richard Maw [2015-09-10 12:13 +0200]: > I've seen this happen when udev is misbehaving because your kernel doesn't > have > CONFIG_FHANDLE enabled. Can you check that first? FHANDLE is enabled on this machine (and Debian jessie in general). --

[systemd-devel] hinting systemd about what to do to make a device appear

Hi, on one of my machines, /srv/backup is an encrypted filesystem sitting on top of a DRBD volume. In /etc/crypttab, I thus have backup /dev/drbd/by-res/backup none noauto and then I mount /dev/mapper/backup on /srv/backuppc. systemd-cryptsetup-generated creates a unit for this which

Re: [systemd-devel] hinting systemd about what to do to make a device appear

On Thu, Sep 10, 2015 at 12:00:50PM +0200, martin f krafft wrote: > it times out waiting for dev-drbd-by\x2dres-backup.device, even > though the file in /dev (and DRBD) are ready to go. I've seen this happen when udev is misbehaving because your kernel doesn't have CONFIG_FHANDLE enabled. Can you

Re: [systemd-devel] hinting systemd about what to do to make a device appear

also sprach Richard Maw [2015.09.10.1227 +0200]: > > FHANDLE is enabled on this machine (and Debian jessie in general). > > I thought I'd check that first, since it has been the root cause for a few > similar issues. It was a good call and I appreciate your time.

Re: [systemd-devel] hinting systemd about what to do to make a device appear

also sprach Lennart Poettering [2015-09-10 13:01 +0200]: > Hmm. After the device appeared in systemd, does it show up in the > Aliases= field of the .device unit? Specifically, can you make the > device appear, then run "systemctl show -a" on it, and paste the > output

Re: [systemd-devel] hinting systemd about what to do to make a device appear

On Thu, 10.09.15 12:33, martin f krafft (madd...@madduck.net) wrote: > also sprach Lennart Poettering [2015-09-10 12:28 > +0200]: > > Does udev know about the symlink? I.e. does "udevadm info" show it? > > Which component is creating it? Only symlinks created by udev are

Re: [systemd-devel] hinting systemd about what to do to make a device appear

2015-09-10 12:00 GMT+02:00 martin f krafft : > Hi, > > on one of my machines, /srv/backup is an encrypted filesystem > sitting on top of a DRBD volume. In /etc/crypttab, I thus have > > backup /dev/drbd/by-res/backup none noauto > > and then I mount /dev/mapper/backup on

Re: [systemd-devel] hinting systemd about what to do to make a device appear

also sprach Michael Biebl [2015-09-10 19:13 +0200]: > Could you run udevadm info on the device? > Does it have a SYSTEMD_READY flag set? No: http://slexy.org/raw/s2VEGDwICN -- @martinkrafft | http://madduck.net/ | http://two.sentenc.es/ murphy's law is recursive. washing

Re: [systemd-devel] hinting systemd about what to do to make a device appear

2015-09-10 19:07 GMT+02:00 martin f krafft : > also sprach Michael Biebl [2015-09-10 17:51 +0200]: >> drbd requires network access and drdb.service has >> DefaultDependencies=yes and is started in multi-user.target. >> >> If this mount point is treated as