SQS has occasional message latencies of up to 5 minutes

0

I have been using SQS (in ap-northeast-1) for years now, but recently I have started noticing within my application logs that a small fraction of my message deliveries have a latencies of up to 5 minutes. I would say 99.9% of the messages are delivered within <1s, but my application is sensitive to these occasional delays. I was wondering what is the AWS SLA for SQS messages latencies once they have been received by SQS to be sent to the registered receiver and if there is a way to minimize the occasional 5 minute latencies on the 0.01% of the messages.

Is moving to AWS MQ or running my own message broker the only solution? I don't really need all the advance features the former provides, I just need SQS to be 100% reliable and consistent when it comes to delivery latency.

kgston
已提問 2 年前檢視次數 1942 次
3 個答案
1

Thank you for the link to the SLA, but it only talks about uptime, for which there was no downtime because the message was received by SQS and eventually delivered, albeit after 5 minutes. I was referring to message latency SLAs. For example SQS agrees to deliver all received messages that has no delay specified within 3s or something to that extent.

I would raise a support case too, except that I'm on the basic support plan and I cannot open technical support cases. I could pay 3% of my billed amount to upgrade my support plan, but I don't really need any technical guidance. I just would like to know if this is a bug/issue on SQS and report it in hopes it would get resolved. I would be happy to provide a few relevant message IDs so a technical support specialist can investigate further.

So far, no errors. No downtime notices on my personal health dashboard and no errors from SQS. I have other messages on other queues being processed and other messages on the same queue coming in just fine.

kgston
已回答 2 年前
0

Are you seeing any errors when the delays occur? FYI the SQS SLA is on our website.

In this case I would raise a support case; talk the team about when this is occurring and try to determine what the root cause is.

profile pictureAWS
專家
已回答 2 年前
0

I agree with @kgston and there is no mention of response time or latency in the SLA link. Amazon talks high about SQS scalability but this doesn't make sense if low latency can't be guaranteed. For several of our SQS usages in Asia Pacific (Mumbai) / ap-south-1 we are experiencing such high latencies (b/w 1 to 5 min) for approx 1% of send requests. About 10% would range in 500ms to 1s and this is when using connection pool. It is desirable to have a guarantee of under 300ms at least.

已回答 1 年前

您尚未登入。 登入 去張貼答案。

一個好的回答可以清楚地回答問題並提供建設性的意見回饋,同時有助於提問者的專業成長。

回答問題指南