Unfortunately we experienced a technical issue which has lead to partial data loss for about half of the uploads made on September 21st and 22nd. Please check your feeds - if you see status “Failed” or “Ingesting” from uploads made on those days (and you don’t see image thumbnails for those uploads), and you still have the original imagery, please go ahead and upload again. Uploads before September 21 and starting from September 23rd are OK (though some may experience processing delays for several days)
We apologize for this technical issue and are taking mitigation steps to try and prevent this sort of thing from happening again in the future.
- Boris, Product Manager at Mapillary
Thank you for acknowledging and properly communicating about this mishap. Although mishaps may entail some levels of inconvenience to others they will eventually happen to everybody. True leadership and accountability is not only about preventing mishaps in the first place but as much about professional mitigation.
We are still experiencing problems processing uploads. Uploads from 23 to 26 have still not been processed.
Thank you for your kind words and understanding @GITNE and for your active participation in the community!
@Lowiekse The pipelines are still catching up from the disruption, but you should be seeing more things being processed and all of them should complete. Send us an email at email@example.com if you don’t see that happening by Monday with a link to your Mapillary profile.
@boris , thank you for updating the community on this issue. I had a series of sequences from September 21 and 22 that I tried reuploading, but the uploader gives an error that these files have already been uploaded. Changing their file names or paths does not work. Is there a solution for this?
Try purging the upload history in
@adalvit @boris Ah btw, the upload history on Linux or XDG compliant systems should go into
$XDG_CACHE_HOME/mapillary_tools instead of
$XDG_DATA_HOME/mapillary_tools because this data is non-essential and can be recreated anytime.
$XDG_DATA_HOME is basically only for user specific shared application data. There is no need nor any reason for sharing a user’s upload history with other apps. Besides, the upload history should have a limited lifetime as opposed to configuration data like a user’s account access token.
Hi @gklafeta, yes, we’ve added an option in the 4.1.3-beta version of desktop uploader to “skip uploaded files” which you can disable by tapping on your username in the top left and selecting “Settings”
You can download the beta version for Windows , Mac & Linux .
Once installed, the beta version will be called “Mapillary Uploader BETA” so you can keep using the public version along with it. The beta will update to the latest version automatically so you’ll always have access to the latest features in development.
cc:@nikola on path comments above
Hi folks, unfortunately we have detected data loss on October 6 which led to about 4% of uploads failing to process. Normally we are able to recover and re-process failed uploads, but in this case we are unable to do so. Again, we apologize for this technical issue and are taking mitigation steps to try and prevent this sort of thing from happening again in the future (including a major re-write of this process to improve reliability).
Please check your feeds - if you see failed sequences for October 6th and have the original imagery, please re-upload following the instructions above.
- Boris - Product Manager, Mapillary
Just October 6? It seems that the following sequences from October 8 were lost:
(I am checking just Berlin area, maybe there’s more lost in the world)
@adalvit - are you able to look at these sequences?
Hi eserte, the images are still there, e.g. the first one you’re linking is here . Since we had to reprocess data from 2023-10-08, a lot of images were assigned new ids. The parts of the process that put sequences together and show them on the map are sill executing, so the sequence you see in blue is a ghost of the old one and is not really functional, and image navigation is also broken. In a few days everything should be restored correctly - have a look on Monday and let’s see if everything is resolved by then.
Thank you for reporting!
OK, now I see the sequences again in the API, but indeed with different sequence IDs than before yesterday.