[pbs-devel] [PATCH backup] fix #2909: handle missing chunks gracefully in garbage collection

Fabian Grünbichler f.gruenbichler at proxmox.com
Wed Aug 5 12:11:37 CEST 2020


On August 5, 2020 6:50 am, Dietmar Maurer wrote:
> But now the task finish with status Ok?
> 
> IMHO, this is a serious problem...
>

for verify this should cause the task to fail, but for GC I am not so 
sure. the garbage collection itself is fine (by definition, any missing 
chunk the GC encounters is one that should not be garbage collected), it 
just found an unrelated problem?

>> On 08/04/2020 4:01 PM Oguz Bektas <o.bektas at proxmox.com> wrote:
>> 
>>  
>> instead of bailing and stopping the entire GC process, log the missing
>> chunk errors and continue.
>> 
>> Signed-off-by: Oguz Bektas <o.bektas at proxmox.com>
>> ---
>>  src/backup/datastore.rs | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>> 
>> diff --git a/src/backup/datastore.rs b/src/backup/datastore.rs
>> index ffd64b81..063fba53 100644
>> --- a/src/backup/datastore.rs
>> +++ b/src/backup/datastore.rs
>> @@ -436,8 +436,8 @@ impl DataStore {
>>              tools::fail_on_shutdown()?;
>>              let digest = index.index_digest(pos).unwrap();
>>              if let Err(err) = self.chunk_store.touch_chunk(digest) {
>> -                bail!("unable to access chunk {}, required by {:?} - {}",
>> -                      proxmox::tools::digest_to_hex(digest), file_name, err);
>> +                worker.log(&format!("warning: unable to access chunk {}, required by {:?} - {}",
>> +                      proxmox::tools::digest_to_hex(digest), file_name, err));
>>              }
>>          }
>>          Ok(())
>> -- 
>> 2.20.1
>> 
>> 
>> _______________________________________________
>> pbs-devel mailing list
>> pbs-devel at lists.proxmox.com
>> https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel
> 
> 
> _______________________________________________
> pbs-devel mailing list
> pbs-devel at lists.proxmox.com
> https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel
> 
> 
> 





More information about the pbs-devel mailing list