[ZFS] Error on exiting Suspend to RAM (partially solved)

[Anard]

Hi.

I just installed a new hard drive to replace older one formatted in HFS+. I installed sys-fs/zfs and made a zpool on this new disk. I copied all my data from older disk and now stopped using older one.
I want to validate this new filesystem before going further. In future, I first want to attach my old disk as a ZFS mirror, and next buy a 3rd disk to use a 3 disks zraid system. This to securize my data and extend my virtual disk capacity without the need of buying very large hard drives.

Anyway. I enconter a problem when my system goes in sleep mode. ZFS sees I/O errors on my drive and stops using it as it can't restore it from another drive. For the moment my zpool contains only 1 drive.
Before erasing my old HFS disk to use it in my ZFS mirror, I'd like to be sure it won't corrupt any data.

Do you know why it doesn't support hibernation and how to solve this problem ?

Thanks for help.
_________________
"iMack" : GA-H97M-D3H, Intel i7 4790, 16Go DDR3, Sapphire RX570 4Go, 2x SSD 256Go, HDD 500Go + Zpool 3x2To / Clover - macOS Mojave / Gentoo-Xfce

[alamahant]

Hibernation means sleep to swap,no?
Do you have a swap partition?
Do you have a resume=UUID=<uuid of swap partition> in grub cmdline?

[Anard]

[Hu]

Anard: you keep referring to hibernation, but as alamahant indicated, "hibernation" is typically understood to mean that the system writes RAM to swap, then turns off power. Later, when power is turned on, a new kernel boots, notices the hibernation image, and copies that into RAM, putting the applications back in the state they were in at hibernate time. Your second post states that you are using suspend-to-RAM, often known as "S3". This is a different power-saving mode, and has different potential failure modes. Therefore, you have a suspend-to-RAM problem, not a hibernation problem.

Your problem looks to me like your drive does not recover after S3, and the kernel passes through that I/O no longer works. Does dmesg after resume agree with that? After this failure, what is the output of dd if=/dev/device-for-zfs of=/dev/null iflag=direct bs=4K count=1?

[Anard]

OK sorry for this approximation (and bad english )

here is :

[Hu]

[Anard]

OK !!
Not completely solved :

Drive was plugged via a SATA extension PCI card (Marvell 9230). This to open the ability of adding drives in future.
As I have a free SATA port on my MB, I just tried to switch the connection of the disk to this free port (PCI SATA expander is not used at all now).
Seems to be OK, but I still have errors in dmesg :

[Hu]

It may be possible to get the Marvell PCI extender to work, but I cannot tell you how to do so, or even if it can be done.

That "slow to respond" is unfortunate, but if everything else works, I think it can be ignored. Is ata14 the Marvell extender? Even though you have no disks on it, if the card is still connected, it might be getting detected and probed, then triggering this warning when the probe fails.

Omitting the error text reproduction is fine, since you lost access to it by fixing the underlying problem.

[Anard]

It's possible. As you said, the extender is still connected, but without disks attached to it.
Maybe should it be possible to ask my system to wait several seconds before trying to wake up these ports ?
Else, I'll try to connect my DVD writer & my /home disk (ext4) and see how it works. Which would free 2 SATA ports from my motherboard.
Anyway, thanks you 2 for help.
_________________
"iMack" : GA-H97M-D3H, Intel i7 4790, 16Go DDR3, Sapphire RX570 4Go, 2x SSD 256Go, HDD 500Go + Zpool 3x2To / Clover - macOS Mojave / Gentoo-Xfce