River Tarnell river@loreley.flyingparchment.org.uk wrote:

Ja Ga:

...

Is it coming back soon? All my stuff is broken, and people are losing patience.

it will be back once the import is finished.

Has this completed yet? I have tools using that database and now I have JIRA requests asking me why my bots aren't updating statistics...

Please advise.

On Tue, Aug 4, 2009 at 12:55, Casey Brownlists@caseybrown.org wrote:

On Tue, Aug 4, 2009 at 4:50 AM, James R.e.wikipedia@gmail.com wrote:

...

Has this completed yet? I have tools using that database and now I have JIRA requests asking me why my bots aren't updating statistics...

http://status.toolserver.org/

-- Casey Brown Cbrown1023

---

Toolserver-l mailing list Toolserver-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/toolserver-l

Sadly the status hasn't changed much in the last days, I think people wonder if you have an clue when it might be done, is it days, weeks or months we are talking about?

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

Andrew Garrett:

Last estimate was about 90 hours.

this estimate is not accurate. i made the comment about 90 hours in the private admins channel on IRC based on a simplistic (and inaccurate) calculation; i expect the real time will be somewhat less than that.

for comparison, 7 hours later, the same simplistic estimate now suggests that import will take 54 hours.

nonetheless, the import is taking longer than i initially expected. i'm sorry for providing a bad estimate, but the import had to be done. it's unfortunate that s1 has to be offline while it's done, but this is a direct result of WM-DE not having enough money to fully support the Toolserver. the recent grant from WMF should help a lot with downtime caused by database issues like this.

under the terms of our SLA, a full refund of account fees for the import period will be given to all users.

- river.

"Carl Fürstenberg" azatoth@gmail.com wrote:

Sadly the status hasn't changed much in the last days, I think people wonder if you have an clue when it might be done, is it days, weeks or months we are talking about?

River Tarnell <river at loreley.flyingparchment.org.uk> wrote at Fri Jul 31 20:42:13 UTC 2009:

River Tarnell:

...

there will be a few hours of downtime while the new database is imported, then some replication lag.

this maintenance has now started; s1 will be offline until it finishes.

> - river.

Clearly, Carl, you are suffering from the same reading comprehension difficulty as me, and other Toolserver users. We were told quite clearly, four and a half days ago, that s1 will be offline for "a few hours." I'm not sure whether it was "few" or "hours" that tripped me up, but it seems certain that at least one of those words means something very different than what I thought it did. I apologize for the misunderstanding.

"Andrew Garrett" agarrett@wikimedia.org

This is totally unproductive, unwelcome and unnecessary. Please go away if this is the sum of your contribution to this list.

You seem to be making sarcastic comments attacking toolserver admins for making an estimate an order of magnitude or two about how long the s1 reimport will take. I'm not sure how you could possibly think this is acceptable.

I should remind you that we can't make the reimport run faster, that we are volunteers (at least, in our capacity as toolserver admins), and that when somebody asked how long was remaining, an up to date estimate was given within ten minutes. What more would you like?

No offense was intended. I was trying to be lighthearted.

I did not see your response before posting. Up until that point, there had been several other inquiries, both on the mailing list and on IRC, and all the responses had been rather curt references to the fact that a reimport was in progress, with no estimate of when it might be completed.

To put it in perspective, I provide professional services for a living. If I told a client that a particular task would be done in a few hours, and several days later it was not done, and if the client then asked what was taking so long, and my response had the tone of the responses that Toolserver users had received *prior to yours* earlier this hour -- I'd be out looking for a new client.

Yes, I know that you are not doing this for pay (although both WMF and Wikimedia DE do actively solicit for monetary contributions from their loyal users), and that you cannot make it go any faster. I think all the users want is to be kept informed if things are taking longer than originally anticipated.

Again, I was not intending to attack anyone, and if my comment was interpreted that way, I do apologize.

Hello, Am Dienstag 04 August 2009 19:16:51 schrieb Aryeh Gregor:

provide not even the vaguest idea of when it will be back up after repeated inquiries.

the truth is that we don't know it. If all goes well, it take a few days (I think River meant "days" insteadt of "hours") to reimport a cluster. Should we say next time "It need 1 month" and then after maybe 4 days "We are back ealier!" (the Scotty-methode) - would that help?

It is realy not the first time that we import a dump - why have the roots to tell that to the new ones, insteadt of the veterans?

To tell a little about the current process: Everthing, but revision, is imported allready. But it can anytime happend that we notice that the dump is corrupt and we have to start at the very beginn.

Another thing: There will be no commons on s1 after the re-import is done for a while - because it have to import seperatly.

Sincerly, DaB.

On Tue, Aug 4, 2009 at 2:31 PM, DaB.WP@daniel.baur4.info wrote:

the truth is that we don't know it. If all goes well, it take a few days (I think River meant "days" insteadt of "hours") to reimport a cluster. Should we say next time "It need 1 month" and then after maybe 4 days "We are back ealier!" (the Scotty-methode) - would that help?

We can give the best estimate we have, and revise it occasionally as things progress. Like "The dump should take something like 2-4 days if all goes well, and potentially much longer if errors occur and it has to be restarted." If the initial estimate turns out to be wrong, we can revise it. In this case we gave a bad estimate and then didn't give revised estimates once the original estimate was clearly wrong. That's an understandable mistake on our part given the givens (volunteers etc.), but it's also understandable that it will annoy users if we forget to give them essential information, and we shouldn't be snapping at them for getting annoyed at our failures.