[pve-devel] [PATCH storage] fix #3203: report smart status correctly

[pve-devel] [PATCH storage] fix #3203: report smart status correctly

[Oguz Bektas]

if the -a option isn't passed, -H might report a failing disk as
'PASSED' even when the disk might be in a corrupted state.

Signed-off-by: Oguz Bektas <o.bektas@proxmox.com>
---
 PVE/Diskmanage.pm | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/PVE/Diskmanage.pm b/PVE/Diskmanage.pm
index 64bb813..d3c3365 100644
--- a/PVE/Diskmanage.pm
+++ b/PVE/Diskmanage.pm
@@ -95,8 +95,9 @@ sub get_smart_data {
 	    or die "failed to get nvme controller device for $disk\n");
     }
 
-    my $cmd = [$SMARTCTL, '-H'];
-    push @$cmd, '-A', '-f', 'brief' if !$healthonly;
+    my $cmd = [$SMARTCTL];
+    push @$cmd, '-H' if $healthonly;
+    push @$cmd, '-a', '-A', '-f', 'brief' if !$healthonly;
     push @$cmd, $disk;
 
     eval {
-- 
2.20.1

Re: [pve-devel] [PATCH storage] fix #3203: report smart status correctly

[Dietmar Maurer]

This code is quite strange. Please can you use a
normal: if .. then .. else ..

> +    push @$cmd, '-H' if $healthonly;
> +    push @$cmd, '-a', '-A', '-f', 'brief' if !$healthonly;

Re: [pve-devel] [PATCH storage] fix #3203: report smart status correctly

[Dominik Csapak]

On 3/1/21 14:56, Oguz Bektas wrote:
> if the -a option isn't passed, -H might report a failing disk as
> 'PASSED' even when the disk might be in a corrupted state.
> 
> Signed-off-by: Oguz Bektas <o.bektas@proxmox.com>
> ---
>   PVE/Diskmanage.pm | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/PVE/Diskmanage.pm b/PVE/Diskmanage.pm
> index 64bb813..d3c3365 100644
> --- a/PVE/Diskmanage.pm
> +++ b/PVE/Diskmanage.pm
> @@ -95,8 +95,9 @@ sub get_smart_data {
>   	    or die "failed to get nvme controller device for $disk\n");
>       }
>   
> -    my $cmd = [$SMARTCTL, '-H'];
> -    push @$cmd, '-A', '-f', 'brief' if !$healthonly;
> +    my $cmd = [$SMARTCTL];
> +    push @$cmd, '-H' if $healthonly;
> +    push @$cmd, '-a', '-A', '-f', 'brief' if !$healthonly;
>       push @$cmd, $disk;
>   
>       eval {
> 

after looking at our code, this patch does not solve the issue the
user in #3203 has

we purposely ignore the return value of smartctl except
for sever smartctl errors (such as device open failed or faulty command 
line) and parse the status from the text itself

but the text (PASSED) does not change regardless if -H is given or not

the real fix for the issue is to start parsing the smart errors 
separately and show them to the user somehow

while doing this, we should change our parser to use the '--json' flag
of smartcl

Re: [pve-devel] [PATCH storage] fix #3203: report smart status correctly

[Oguz Bektas]

hi,

On Tue, Mar 02, 2021 at 09:34:41AM +0100, Dominik Csapak wrote:
> On 3/1/21 14:56, Oguz Bektas wrote:
> > if the -a option isn't passed, -H might report a failing disk as
> > 'PASSED' even when the disk might be in a corrupted state.
> > 
> > Signed-off-by: Oguz Bektas <o.bektas@proxmox.com>
> > ---
> >   PVE/Diskmanage.pm | 5 +++--
> >   1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/PVE/Diskmanage.pm b/PVE/Diskmanage.pm
> > index 64bb813..d3c3365 100644
> > --- a/PVE/Diskmanage.pm
> > +++ b/PVE/Diskmanage.pm
> > @@ -95,8 +95,9 @@ sub get_smart_data {
> >   	    or die "failed to get nvme controller device for $disk\n");
> >       }
> > -    my $cmd = [$SMARTCTL, '-H'];
> > -    push @$cmd, '-A', '-f', 'brief' if !$healthonly;
> > +    my $cmd = [$SMARTCTL];
> > +    push @$cmd, '-H' if $healthonly;
> > +    push @$cmd, '-a', '-A', '-f', 'brief' if !$healthonly;
> >       push @$cmd, $disk;
> >       eval {
> > 
> 
> after looking at our code, this patch does not solve the issue the
> user in #3203 has
> 
> we purposely ignore the return value of smartctl except
> for sever smartctl errors (such as device open failed or faulty command
> line) and parse the status from the text itself
> 
> but the text (PASSED) does not change regardless if -H is given or not
> 
> the real fix for the issue is to start parsing the smart errors separately
> and show them to the user somehow

hmm yeah you're right, thanks!

> 
> while doing this, we should change our parser to use the '--json' flag
> of smartcl

i will take a look. like you also mentioned off-list we do this in
backup server code with the json flag, so it makes sense to do it here
as well

> 
> 
> _______________________________________________
> pve-devel mailing list
> pve-devel@lists.proxmox.com
> https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
> 
>