Re: [PATCH proxmox-backup v3 3/3] fix #7400: proxy: self-heal corrupted job statefiles

["Michael Köppl" <m.koeppl@proxmox.com>]

On Thu Apr 2, 2026 at 6:50 PM CEST, Christian Ebner wrote:
> One nit inline.
>
> Other than that:
>
> Reviewed-by: Christian Ebner <c.ebner@proxmox.com>
>
> On 3/25/26 5:06 PM, Michael Köppl wrote:
>> Update update_job_last_run_time to transition JobState::Unknown into
>> JobState::Created so the corrupted statefile is overwritten. In
>> addition, update the scheduling loops to actively overwrite corrupted
>> statefiles and return the time for the next scheduled run of the
>> affected job.
>> 
>> Signed-off-by: Michael Köppl <m.koeppl@proxmox.com>
>> ---
>>   src/bin/proxmox-backup-proxy.rs | 4 +++-
>>   src/server/jobstate.rs          | 2 +-
>>   2 files changed, 4 insertions(+), 2 deletions(-)
>> 
>> diff --git a/src/bin/proxmox-backup-proxy.rs b/src/bin/proxmox-backup-proxy.rs
>> index c1fe3ac15..71d8566f4 100644
>> --- a/src/bin/proxmox-backup-proxy.rs
>> +++ b/src/bin/proxmox-backup-proxy.rs
>> @@ -564,6 +564,7 @@ async fn schedule_datastore_garbage_collection() {
>>               Ok(time) => time,
>>               Err(err) => {
>>                   eprintln!("could not get last run time of {worker_type} {store}: {err}");
>> +                let _ = jobstate::update_job_last_run_time(worker_type, &store);
>>                   continue;
>>               }
>>           };
>> @@ -936,7 +937,8 @@ fn check_schedule(worker_type: &str, event_str: &str, id: &str) -> bool {
>>           Ok(time) => time,
>>           Err(err) => {
>>               eprintln!("could not get last run time of {worker_type} {id}: {err}");
>> -            return false;
>> +            let _ = jobstate::update_job_last_run_time(worker_type, id);
>> +            proxmox_time::epoch_i64() - 30
>
> nit: Should we define the offset from epoch here as constant? This is 
> used in several places now and having one common constant with a 
> meaningful name would help with code quality IMHO.

Thanks for a having another look at this version! I agree, that's a good
idea. What do you think of SCHEDULE_FALLBACK_OFFSET?

>
>>           }
>>       };
>>   
>> diff --git a/src/server/jobstate.rs b/src/server/jobstate.rs
>> index 4163656e8..214691965 100644
>> --- a/src/server/jobstate.rs
>> +++ b/src/server/jobstate.rs
>> @@ -156,7 +156,7 @@ pub fn update_job_last_run_time(jobtype: &str, jobname: &str) -> Result<(), Erro
>>               state,
>>               updated: Some(time),
>>           },
>> -        JobState::Unknown => bail!("cannot update last run time for unknown job state"),
>> +        JobState::Unknown => JobState::Created { time },
>>       };
>>       job.write_state()
>>   }