Twitter | Search | |
Mark Brown
I work at Microsoft on the Azure Cosmos DB team. Love distributed computing and testing CAP/PACELC theorem.
9,520
Tweets
599
Following
2,365
Followers
Tweets
Mark Brown 5h
Reply Retweet Like
Mark Brown Aug 10
btw, for Mongo you will need to manage your own failover as I don't believe their driver includes automatic failover logic.
Reply Retweet Like
Mark Brown Aug 10
No, it's twice. Single-master = .08c/hour per 100 RU/s.
Reply Retweet Like
Mark Brown Aug 9
.016c/hour per 100 RU/s Pricing page here Capacity Calculator is helpful too.
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
No problem. Happy to help.
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
Well if you had multi-region-writes enabled you potentially could retry the write in another region and then the service would deal with it as it worked to replicate to the primary region but that seems a bit of over kill. Replicas don't fail often so I wouldn't worry.
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
ok, so yeah that was likely service retrying which is why the latency was so high and is only temporary. When a replica fails there may be some latency while the set reconfigures and brings up a new replica.
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
Looks like one of the clusters is having an issue. Is this still occurring? A single cluster shouldn't effect availability since we run 4 replica per replica set across fault domains.
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
Let me take a look at our service telemetry.
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
I've not used App Insights on my Cosmos apps. "Faulted" is useless as a result. But the time it took to fail suggests either a service issue or request latency from retry's due to 429's. Are you inserting a large amount of data?
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
So writes are taking 30 seconds and then failing?
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
Are you logging errors back from our SDK? What's the HTTP Status Code?
Reply Retweet Like
Mark Brown Aug 9
Replying to @flying_maverick
What is that a screen cap of?
Reply Retweet Like
Mark Brown Aug 9
When we release in preview it will have some limits on it to start. Notably just single region. But will open when we go GA.
Reply Retweet Like
Mark Brown Aug 9
Each region requires its own compute (throughput). Free tier with 400 RU/s would only cover running in a single region.
Reply Retweet Like
Mark Brown Aug 6
Not at the moment. If you have any you'd like to see please add on User Voice.
Reply Retweet Like
Mark Brown Aug 6
Spatial Functions are here.
Reply Retweet Like
Mark Brown Aug 6
Replying to @ssmsexe @AzureCosmosDB
Actively working on that. Shooting for before end of year but don't hold me to it. ;)
Reply Retweet Like
Mark Brown Aug 5
Replying to @tomasrestrepo @oising
Would emulator work for your test environment?
Reply Retweet Like
Mark Brown Aug 5
Replying to @tomasrestrepo @oising
Yes, that is a gap on the min RU/GB. We do now include this in our Capacity Calculator and in our docs on Cost we specifically tell folks to use the calculator to determine total storage cost.
Reply Retweet Like