@neamanshafiq and all
The WTT update did run this morning (Sunday) but stopped again. This time much further on, at N15.
So that should mean that day services for P4 and above are missing (including river and tube replacement services). Also night services on N2-N9, N16-N98, N155-N551 and on 24 hour routes 2-6, 23-94 and 159 upwards. So a majority rather than (as last week) minority of files are present.
The point at which it throws a wobbly seems to be random, which is not going to help in tracking the fault down.
@neamanshafiq and all
Appear to have a complete set of WTTs again this morning but unless what is happening is understood the issue is likely to recur, I fear.
Just in case it is relevant, I observe that only the files missing yesterday are dated 12th. The others are still dated 11th (in a few cases earlier) so it looks like the rerun was not recreating those files again from scratch. Might that mean that it is the file creation rather than transfer procedure which is falling over?
Thanks @mjcarchive - I’ve added this update to the ticket I’ve opened with our internal support team. I’m going to ask a separate team to check the total number of files transferred on 11th and 12th and whether the file transfer only pushes up any new/changed files.
@neamanshafiq
The file transfer usually overwrites everything - except of course when there is no new file to replace the old. It would actually make life easier if it did just push up genuinely new files, as just the new ones could be downloaded by users, but clearly the files themselves are recreated each week (often with tiny variations in size) so checking each file for newness would be an extra step.
At the moment there are 3041 WTTs listed in the data bucket.
748 of these are dated 12th
2289 are dated 11th
4 are dated 4th (they have been replaced by different day types, thus not overwritten and will no doubt be deleted shortly)
And in my case a specific New Years Resolution to try to find and fix the serious fault reported in March at Buses that disappear from predictions API which makes the API, journey planner and countdown screens totally unfit for their purpose, at least in respect of the first half dozen stops for buses leaving the Chingford terminus.
Maybe we’ll both get them, but it will probably be for 2026 or 2028.
@neamanshafiq
The update appears to have been completed this morning.
I don’t think it is complete on the seasonals, as a lot of routes which had Ce WTTs last year do not have them yet this year, including some like route 1 where the timetable very obviously differs from Sa in the latter half of the day. If it appears next week it will be too late to be of other than historical interest! (Also see other thread.)
But that is not an issue about the update process itself.
I wasn’t expecting a Christmas Day update! It fell over after H9 this time by the looks of it as files after that are all dated 26th.
It can’t be something daft like the maximum number of files being exceeded. Can it? IIRC that was the cause of the Datastore JP files update falling over a few years back.
@neamanshafiq
This week the process (whichever is at fault) has stopped in the middle of doing route 8, thus slightly further than last week’s first effort.
If it went on a bit further so as to go “99, bonk” I could at least resurrect the centipede joke from my schooldays.
@neamanshafiq
The missing stuff completed this morning.
One point worth noting. Since the upload moved to Sunday, we only get the Sa and Su WTTs were there has been a service change, the MF following the next weekend. The 142 and 187 are just two of the examples this week. It is not universal as (for example) all files for the 218 and 226 are present.
The logic (for the 142 etc.) would seem to be that a new WTT should only be uploaded once it has actually been operated (which on the Sunday it would not have been). For the 218 etc. it would seem to be that a new WTT should be uploaded once the WTT it is replacing has gone dead (which an obsolescent MF schedule would have done by Sunday). Not sure why it varies (different operators?) but clearly the second (218 style) approach is better for the user.
I also note that five files dated 28/01/2023 have been loaded (routes 35 and 176). Either they are dated wrongly or they (or at least the 176 files; the 35 is a “special”) have prematurely overwritten the current files. So that is a third variant of timetabling updates!
@mjcarchive I’ve been informed by the support team that they have deployed a change for a related issue and have asked if we can check over the next few days to see if that has resolved our issue. I’ll be checking the number of files transferred in total over the next few days to see if there has been any improvement.
@neamanshafiq
Thanks. “Related issues” always pique my interest! It’s usually the case that the error you see from a computer system is a symptom of a bug several thousand lines of code earlier…
@neamanshafiq
… and completed this morning. Any ideas yet? I don’t think it can be a limit on the number of WTTs that can be handled as there are scarcely any more this week than last but it didn’t get anywhere near as far. A limit on the overall number of files that can be bucketed? Seems unlikely.
Do we even know whether it is the file creation or the file uploading aspect which is crashing?
@mjcarchive still under investigation I’m afraid. it feels like there is a pattern here where only some files are pushed through on sunday morning and the rest are pushed through the next morning. would you agree?
@neamanshafiq
Absolutely - I thought I had been saying that for weeks!
Thinking about it, given that the entire set is supposed to be updated in one go, the fact that it picks up the next morning where it left off rather than starting from scratch (and failing again) is quite interesting, as it suggests that the rest of the job has been left queued for the next overnight window. In the interests of ruling out the simplest explanations first, is that window perhaps too short to enable the Sunday morning upload to complete? (The even simpler explanation, that it will only do up to N files, does not seem to fit the evidence.)