Access current maxlag value of the server

[tfrancart]

A query with -1 as the maxLag value returns the actual server lag without any metadata.
I would find it convenient if the API could provide a simple "ping" call that would return the current server maxLag, so that we are informed how busy the server is. I could not find a way to know this besides receiving an Exception when the call fails due to maximum number of retries reached.

[tfrancart]

e.g. https://www.wikidata.org/w/api.php?action=query&titles=MediaWiki&format=json&maxlag=-1

[wetneb]

The code for this is in #470, hold my beer

[tfrancart]

Cool. The throttling parameter the PR contain might very well explain why I am receiving tons of "[failed-save] The save has failed." Exceptions because edits are performed too rapidly. Do I understand correctly Wikibase refuses updates if the lag is too high ?

[tfrancart]

Argh, just saw you removed that throttling logic thing, I could have used that, too.
What would be your recommendations to avoid [failed-save] Exceptions ? pause before each call ? ask for maxLag before each call and if above a threshold, pause ?

[wetneb]

Yes, by default WDTK will send a maxlag=5 parameter in all its editing requests, so that they fail if the lag is above 5 (which is the officially recommended value for this parameter on Wikidata).

If you want your bot to be more aggressive (keep editing even if the lag is high) you can use setMaxLag() on the ApiConnection to change this to a higher value, and your bot will keep running even if the servers are behind.

Concerning the throttling, we lack official guidance from WMDE to implement this properly. https://phabricator.wikimedia.org/T240442

[tfrancart]

I am seeing 2 kinds of different behaviors :

1. Sometimes the API call just hungs, in that case I know it is in a retry loop because maxLag is too high, and if after 14 retries lag is till too high, I get a MaxlagErrorException: [maxlag] Waiting for all: 10.016666666667 seconds lagged.

2. But sometimes after the bot has made 7 or 8 edits, I get a bunch of [failed-save] The save has failed. Exception directly, without waiting for any retry, and then processing resumes as normal, again for 7 or 8 edits, and then a bunch of Exceptions, etc. Is this second behavior also driven my maxLag ? should I try increasing averageTimePerEdit maybe ?

[wetneb]

1. Yes the lag can stay above 5 for hours and hours…
2. Perhaps linked to https://phabricator.wikimedia.org/T255804?

Overall, as client you do not have many possibilities: keep retrying.