I have a bot called Interwicket which keeps itself busy adding and updating the language links for the wiktionaries (namespace 0); it is much more efficient for that than the "standard" pedia bot. It uses the API, and the maxlag parameter, and is a good citizen, backing off sharply as maxlag exceeds 5 seconds, and effectively stopping when it stays above that, until it is again around 5.
Yesterday the reported server lag suddenly became 9000+ seconds, and then slowly ran down, at about 200 sec/min, until it was normal again. A little while ago it was suddenly 80K+, and is now 40K-something. Now I understand why it runs down as something catches up, and why that is faster than "real time".
What I don't get is how the server lag can be < 5 seconds on one request, and 9000 (or 80,000 :-)) on the next request a few seconds later? Could someone tell me what does this; what is happening?
Robert (presently 45971, bot waits patiently ;-)
On Tue, Mar 3, 2009 at 7:02 PM, Robert Ullmann rlullmann@gmail.com wrote:
What I don't get is how the server lag can be < 5 seconds on one request, and 9000 (or 80,000 :-)) on the next request a few seconds later? Could someone tell me what does this; what is happening?
A server was taken out of rotation and its slave process stopped to produce a dump for the toolserver. It was put back into rotation before it caught up, and so it was suddenly the most lagged server in rotation.
Ah, that makes sense. Also explains why the lag dropped back from 42000 to <5 a few minutes ago ;-) Thanks, Robert
A server was taken out of rotation and its slave process stopped to produce a dump for the toolserver. It was put back into rotation before it caught up, and so it was suddenly the most lagged server in rotation.
now we have server lag at 870 seconds, slowly, but inexorably, increasing (?) Robert
On Tue, Mar 3, 2009 at 11:30 AM, Robert Ullmann rlullmann@gmail.com wrote:
Ah, that makes sense. Also explains why the lag dropped back from 42000 to <5 a few minutes ago ;-) Thanks, Robert
A server was taken out of rotation and its slave process stopped to produce a dump for the toolserver. It was put back into rotation before it caught up, and so it was suddenly the most lagged server in rotation.
now going down slowly, as you would expect. Perhaps "normal" ;-)
On Wed, Mar 4, 2009 at 2:07 AM, Robert Ullmann rlullmann@gmail.com wrote:
now we have server lag at 870 seconds, slowly, but inexorably, increasing (?) Robert
On Tue, Mar 3, 2009 at 11:30 AM, Robert Ullmann rlullmann@gmail.com wrote:
Ah, that makes sense. Also explains why the lag dropped back from 42000 to <5 a few minutes ago ;-) Thanks, Robert
A server was taken out of rotation and its slave process stopped to produce a dump for the toolserver. It was put back into rotation before it caught up, and so it was suddenly the most lagged server in rotation.
wikitech-l@lists.wikimedia.org