count buffer/cache into MemInfo

[critical27]

Total = Used + Free + Buffer/Cache

Maybe need to take buffer/cache into consideration, here a screenshot when I do some test.

Graph will keep logging something like Used memory(211303932KB) hits the high watermark(0.800000) of total system memory(263856036KB)..

PR is still in verification.

[bright-starry-sky]

LGTM

[panda-sheep]

Good!

[yixinglu]

We should count the physical memory usage, not include buffer. Here freeInKB() returns the same result as the free column of free -h.

More information refer to https://stackoverflow.com/questions/63166/how-to-determine-cpu-and-memory-consumption-from-inside-a-process

[critical27]

We should count the physical memory usage, not include buffer. Here freeInKB() returns the same result as the free column of free -h.

More information refer to https://stackoverflow.com/questions/63166/how-to-determine-cpu-and-memory-consumption-from-inside-a-process

Could have discuss later. Some known issues when I write data by importer:

1. Failed to write data once pressure if enough
2. Some interface could bypass the check later, show hosts will fail as well if high watermark is reached.

If take buffer into consideration, it will solve these problems.

[yixinglu]

We should count the physical memory usage, not include buffer. Here freeInKB() returns the same result as the free column of free -h.
More information refer to https://stackoverflow.com/questions/63166/how-to-determine-cpu-and-memory-consumption-from-inside-a-process

Could have discuss later. Some known issues when I write data by importer:

1. Failed to write data once pressure if enough
2. Some interface could bypass the check later, show hosts will fail as well if high watermark is reached.

If take buffer into consideration, it will solve these problems.

Good point for question 1. Now you can update this config to 1.0 before importing and restore the config when finishing data loading.

In query engine, we have ignored the checks for most executors except query related executors and if show hosts still report this error, that's a bug.

[CLAassistant]

All committers have signed the CLA.