On 9 May 2017 at 07:18, Rhythmic Fistman <[email protected]> wrote:
>
>
> Il giorno domenica 7 maggio 2017 03:38:18 UTC+10, mpl ha scritto:
>>
>> On 6 May 2017 at 03:05, Rhythmic Fistman <[email protected]> wrote:
>> >
>> >
>> > Il giorno venerdì 5 maggio 2017 23:27:55 UTC+10, mpl ha scritto:
>> >>
>> >> On 5 May 2017 at 09:00, Rhythmic Fistman <[email protected]> wrote:
>> >> > I'm in a bind with this - I ctrl-C camlistored and the kvIndexFile
>> >> > became
>> >> > corrupted (how do I cleanly stop it?) & re-indexing is taking days.
>> >>
>> >> Hmm, how come you were on kvIndexFile? I thought we had made leveldb
>> >> the default everywhere (because we indeed experienced corruption
>> >> problems with kv iirc). Did you specify kvIndexFile yourself on
>> >> purpose?
>> >
>> >
>> > Not on purpose, it's because I come back to camlistore every year or two
>> > to
>> > try and backup and deduplicate my photos,
>> > so I've got old configurations lying around.
>> >
>> > So I switched to levelDB. On my local ssd setup
>> >
>> > $ time camlistored -openbrowser=false -reindex
>> >
>> >
>> > Starting camlistored version 8f1a7df176de21d0d052c04db45b373c5c282fb1;
>> > Go
>> > go1.8 (darwin/amd64)
>> >
>> > Starting to listen on http://localhost:3179
>> >
>> > Index wiped.
>> >
>> > Error: blobpacked storage detects non-zero packed zips, but no metadata.
>> > Please re-start in recovery mode with -recovery.
>> >
>> >
>> >
>> > $ time camlistored -openbrowser=false -recovery
>> >
>> > Starting camlistored version 8f1a7df176de21d0d052c04db45b373c5c282fb1;
>> > Go
>> > go1.8 (darwin/amd64)
>> >
>> > Starting to listen on http://localhost:3179
>> >
>> > ui: serving Closure from embedded resources
>> >
>> > Starting recovery of blobpacked index
>> >
>> > Caught panic installer handlers: error instantiating storage for prefix
>> > "/bs/", type "blobpacked": Sum of all zips (1055150 bytes) does not
>> > match
>> > manifest's WholeSize (527575 bytes) f
>> >
>> > [...] BACKTRACE ELIDED
>> >
>> >
>> > This time reindex works in ~13 minutes
>>
>> wait, are you saying that you got a panic with serverinit, but the
>> camlistored process didn't die, and the reindexing kept going
>> successfully?
>
>
> no - -reindex failed (and exited), then -recovery failed (and exited). After
> that -reindex worked.

oh, that's somewhat reassuring then. But if -recovery was not
successful, you might still have "invisible" blobs. I'll look into it
get back to you.

>> > time camlistored -openbrowser=false -reindex
>> >
>> >
>> > but should I be worried that the recovery failed?
>>
>> Not worried that something is corrupted, but worried that not all your
>> blobs are reachable (because the blobpacked index is not up to date),
>> yes.
>
>
> What should I do?

You could try restarting with -recovery again.

>> Can you please file a bug about it, so I don't forget to try and
>> reproduce/fix it?
>
>
> Ok - not sure how. Maybe I'll back up the index, then try re-indexing.

You would usually go to
https://github.com/camlistore/camlistore/issues and create a new
issue. But nevermind for this time, I've done it already:
https://github.com/camlistore/camlistore/issues/926

>> > similar story with the larger external non-ssd.
>> >
>> > recovery fails, now re-indexing, looks like it might take an hour - this
>> > is
>> > an improvement over kvindex I think.
>> >
>> > Strangely, local SSD had camlistored saturating the CPU (200%?) whereas
>> > external rust is sitting at 6%...
>>
>> yeah I think that's pretty common, hence why I mentioned using an SSD
>> before. afair it's partly due to HDDs having to wait for fsync or
>> something like that.
>
>
> I started over with SSD and the difference was like night and day. I
> processed 200GB of photos in ~1hour, all the cores were saturated.

Cool.

> It didn't all go smoothly though, near the end camlistored hit the open
> files limit and I had to up the ulimit -n from 255 (4096 worked) and start
> over, which was even quicker.

Ugh. afair we fixed some parts of camlistored so they would set the
ulimit for you, but apparently not in all cases yet.

> --
> You received this message because you are subscribed to the Google Groups
> "Camlistore" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"Camlistore" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to