piespy said:
Everything between id:775073..775217 gone.
What the fuck? There's over a one hour gap between these posts, so I highly doubt only spam was uploaded in that hour. Who deleted these posts?
Posted under General
piespy said:
Everything between id:775073..775217 gone.
What the fuck? There's over a one hour gap between these posts, so I highly doubt only spam was uploaded in that hour. Who deleted these posts?
evazion said:
What the fuck? There's over a one hour gap between these posts, so I highly doubt only spam was uploaded in that hour. Who deleted these posts?
A one hour gap between uploads isn't too unusual, especially if the DB is acting up during that time.
I'm not sure how the post counter is implemented, but if it's bumped using a DB query, and that query succeeds but the actual upload doesn't, you'd get this symptom. However, if it's like the mysql auto_increment or using a transaction, then DB atomicity should prevent such a thing.
FWIW, iqdb is grabbing all new posts at 10 minute intervals, and it never got any posts between those IDs either. Neither were there any query failures during that time. So I don't think they were there and then subsequently deleted, instead they're just ghost IDs without ever having a corresponding DB entry. I don't think we lost anything except a small part of ID space.
piespy said:
I'm not sure how the post counter is implemented, but if it's bumped using a DB query, and that query succeeds but the actual upload doesn't, you'd get this symptom.
Hmm, looking into it, this seems to be the issue. The post id counter is incremented on inserts to the post table via a postgresql sequence, but postgres doesn't rollback the sequence increment if the insert fails*. So post id's aren't guaranteed to actually be sequential.
[*]: See the note on the bottom of this page.
evazion said:
Hmm, looking into it, this seems to be the issue. The post id counter is incremented on inserts to the post table via a postgresql sequence, but postgres doesn't rollback the sequence increment if the insert fails*. So post id's aren't guaranteed to actually be sequential.[*]: See the note on the bottom of this page.
I don't understand a word you're saying. I'm just going to assume you mean the chickens that run danbooru ("nobody here but us chickens!") with poultry magic aren't doing their job. Albert, stop being cheep and pay them their chicken feed.
jokes aside, thanks to everybody who're knocking heads to get this fixed.
Problem fixed just now, so it lasted around 2 hours.
Log said:
I wasn't really uploading normally, I got a failbooru after each upload but they went through fine so I ignored them and kept going.
That's strange, since I did retry many times but could not upload anything. On the other hand, error message with editing tag is just for show again.
Other weirdness includes post #776934, thumbnail and sample image showed up just fine, but full view is corrupted at the bottom. I tried to re-upload that one from pixiv source and from my saved file but got duplicate error.
Updated by rantuyetmai
rantuyetmai said:
Other weirdness includes post #776934, thumbnail and sample image showed up just fine, but full view is corrupted at the bottom. I tried to re-upload that one from pixiv source and from my saved file but got duplicate error.
I really, really, REALLY wished you Janitors had the power to double delete post when oddities like that happens. I encounter Janitors more than I encounter Mods and when something like that happens you have to wait a good while before anyone do something about it.
RaisingK said:
Ha, even the MD5 hash is the same as the original.
Wait...so is it corrupted from a file standpoint or it just looks corrupted?
Mr_GT said:
I really, really, REALLY wished you Janitors had the power to double delete post when oddities like that happens.
The fewer the people who can double-delete, the better. Things like this don't happen that often, and it's not a big deal if that faulty post stays up an extra day or two.
Mr_GT said:
Wait...so is it corrupted from a file standpoint or it just looks corrupted?
I meant that the MD5 hash of the image on pixiv is equal to the one Danbooru has assigned to that post (which doesn't match the file Danbooru actually has).
Updated by RaisingK
RaisingK said:
The fewer the people who can double-delete, the better.
While in theory that's reasonable it sorta sounds like the higher up don't trust their janitors enough with something like the ability to double delete post.
Things like this don't happen that often, and it's not a big deal if that faulty post stays up an extra day or two.
The thing that's happening to post #776934 I can agree it doesn't happen too much to be completely worried about, but I seen a good amount of another corruptions happen from time to time like post #776211 (might need to look at it from the status:deleted page to understand what I mean)
I meant that the MD5 hash of the image on pixiv is equal to the one Danbooru has assigned to that post (which doesn't match the file Danbooru actually has).
So it's corrupted but Danbooru see it as the same as the one on Pixiv which is why it's prevent us from re-uploading it. Always wondered what causes these kind of thing anyway.
Mr_GT said:
Not sure if this relates to the continuing failbooru problem or not but has anyone having problem viewing the first post page? All I can see is nothing and not all the post are tagged with loli on them.
I believe the issue behind this was a css error associated with a post that somehow made it's way up with no tags (including tagme.) This occurred the other day on 3dbooru and tagging the image fixed it.
I keep getting failbooru trying to browse my favorites. In fact, now I can't go more than 4 pages if I search my favorites with a specific tag (ex. barefoot)
Updated by ZeetherKID77