URLs (aka web addresses) of any relevant observations or pages: n/a (observations failed to upload)
Screenshots of what you are seeing: nothing - observations are missing
Description of problem (please provide a set of steps we can use to replicate the issue, and make as many as you need.):
Step 1: Used the upload link on the website (https://www.inaturalist.org/observations/upload) to prepare a batch upload (added 72 pictures), combining photos, editing locations, adding notes, tags, and observation fields.
Step 2: Uploaded the observations (27 total). This was the batch upload I did at 9:06 AM EDT this morning.
Step 3: Went back through my observation list to add annotations for plant phenology, noticed that only 25 observations uploaded - 2 observations (8 photos) were missing. Waited ~1 hour to check again in case theyāre taking a while to come through and theyāre still missing.
This seems somewhat unusual website behavior, especially since the other 25 observations uploaded just fine and I canāt think of anything that might have been different about the two that didnāt. I can upload again but donāt want to end up duplicating effort if these are still in limbo somewhere.
I have had this a few times lately when uploading via the website. One ob out of a batch just vanishes without any warning or error. I figured out that it only ever happened in cases where two of the observations were based off the same photo (like if a photo has both a swan and a duck in it, say). Itās happened at least four or five times to me but I havenāt been able to reproduce the error reliablyā¦ It doesnāt always happen when I have a duplicated pair of observations like that, just sometimes, and I have yet to fathom out more precisely the conditions under which it occurs.
Iāve uploaded a lot of observations over the past few weeks, using Firefox, with some duplicated in the uploader, but I donāt think Iāve noticed any that have been lostā¦ It would be difficult to check now, though.
I had this every now and then as well in the last weeks (Browser Firefox). I did not have the feeling that duplicated observations were affected, but am not sure. Maybe also observations that were last (or 2nd last) in the upload line?
Chrome. I should clarify that this wasnāt using the Duplicate button in the uploader, but rather manually creating two observations that had a photo in common (usually as well as at least one unique one).
I did wonder if it was something to do with dragging and dropping the same photo twice onto the uploader. Maybe when it gets to uploading the observation with the second instance of the same photo, something in the system says āoh this photo has already been uploaded successfully so this can be skippedā. Some sort of internal flag marking that the first photo has been uploaded also gets applied to the second instance because it has the same filename, or something like that. That was my best guess at what is happening, but since it doesnāt happen in all cases there is clearly more to it.
None of mine were duplicates, nor were the missing observations the first or last in the batch. It seemed pretty random and I canāt think of any pattern/reason why these two in particular didnāt make it. Interesting to see that others are experiencing similar issues - I always wonder if it is my computer/internet connection or something Iām doing. Iāll have more stuff to upload so Iāll keep an eye on it and see if it happens again.
Yes, like that. 99% of my drag-n-drops are from the regular Windows Explorer window, but I canāt say for certain it wasnāt from the FastStone window so maybe thatās the issue.
can you tell which of the observations that did get loaded would have been the ones immediately before and after the ones that did not get loaded? (for example, missing obs A should have been loaded between obs E and obs F, and missing obs B should have been loaded between obs P and obs Q.)
@pleary searched our logs for your upload, @annkatrinrose. He says 27 observations were created but 2 no longer exist and heās now looking into how that happened.
itās strange to me that two observations would have been created and no longer exist but that their photo records would still exist. just based on my recollection of some testing iāve done in the past, i donāt think thereās a way through the API to delete observations without also deleting the observation records associated with them (if they are tied only to the deleted observations), unless you first disassociate the photo records from the observations before deleting the observations.
if i didnāt know anything else, it would make more sense to me that the system ran into some sort of error while creating the observation records but never reported the issue back to your upload page somehow. but i guess pleary and iNat staff will have a better understanding, as they can see the logs, etcā¦
i donāt see anything unusual about these photo records or any of the observation records in the uploaded set. in terms of the sequence of the missing records, they would have been second and third in the set of 27, if ordered by their observation IDs.
Yep, those photos look about right. There should have been two more for 079. Very strange that the photos would be there but no associated observations.
iNaturalist is likely a highly distributed and asynchronous system. This means that immediate consistency of data cannot be guaranteed, but it tries to reach āeventual consistencyā.
When something goes wrong during upload, several steps may have already completed with succes. In case of an error, retries are expected to be done, and if the error lasts, the distributed transaction has to be rolled back by executing correcting actions, and notifying the client application.
Things do get terribly complicated under such circumstances, and somewhere there the issue has occuredā¦