Get hands-on experience with 20+ free Google Cloud products and $300 in free credit for new customers.

AlloyDB for postgres: Minimum available memory drastically went down

We are running multiple instances of alloy-db for postgres. On may 21, 10PM IST, we started getting alerts for all instances that minimum available memory is very low. I invetigated all services using alloy-db and found that my services are handling normal workload as before. I also checked other metrics for all db instaces like CPU usage, active connection etc. and found no anomaly.

After investigating gcp document for alloy-db. I found that default values for shared_buffer and effective_cache_size are set at 80%. We are also not setting these values from our side and leaving to default values. Also this page is updated recently. So I am concluding right now that this change to 80% might be done recently and that's why we are getting these alerts for all database instaces at same time. However I have not got any prior notice or missed this anouncement from GCP, so I am not fully confident.

Please help investigate this issue. attaching screenshot.min_avail_mem.png

0 2 167
2 REPLIES 2

Hi @brown_dynamite,

Based on the release notes, there haven't been any recent changes specifically related to shared_buffer and effective_cache_size for AlloyDB. In fact, the 80% value you're seeing is the default setting for these parameters, as outlined in the Supported database flags.

It seems like the low memory alerts might be caused by something else, like changes in workload or scaling. Since these settings are default and not something that has changed recently, I’d recommend reaching out to Google Cloud Support for further assistance in troubleshooting your issue. They can help check if anything else is affecting your memory usage or configurations.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.

 

Hi @mcbsalceda , thank you very much for replying. Interestingly, all databases have recovered. We'll look into our workload and try to figure out what went wrong.