Insert failure: why am I not able to insert records?

1

Sep 02 2015 07:53:21 GMT: CRITICAL (namespace): (namespace_warm.c:as_namespace_xmem_delete:499) ns x fail persistent memory delete

this is namespace config

namespace x {
        replication-factor 2
        memory-size 10G
        default-ttl 1000D
        high-water-memory-pct 95
        stop-writes-pct 90
        storage-engine memory
}

Thanks,

2

Hi mwang-

Thank you for writing.

Would you provide more context of the problem? If Aerospike returns a critical message, I expect that the node has stopped. If the node has stopped and you have one node, you cannot insert anything.

I would like to know why the node stopped. I have some guesses based on the one log line, but they are just guesses.

Would you send more details from the logs about the failure?

Do you see anything about the insert failures on the client side? Are you using an API for the client? Are you using aql or cli?

Thank you for your time,

-DM

3

Client is based on GO API. The node did not stop because I can insert some records by cli without critical message. Can you tell me the guesses? Thanks,

Because this is local test, lots of other APIs are running at the same server. So the reason is memory usage on this server hits 90%?

The default memory-size is 10G, while the local data is very small and the aerospike’s memory usage is less than 1%.

4

Hi mwang,

My guess is that the node was trying to warm-start, but that the service was restarted as another user that does not have permissions on the shared memory reserved by the first user, and that is why you see the error.

If the node is alive (and by ‘alive’, I mean that the ‘cake’ message has been written to aerospike.log), can you still reproduce the same issue?

Do you see different results in the api versus cli?

Thank you for your time,

-DM

5

Why this node was trying to warm-start?

6

Hi mwang,

The short answer about warm restarts is that is the default behavior. When an Enterprise node restarts, it attempts to warm start by default. You can read more information about this here.

As for why it would need to restart, I can’t say. The log line that you sent indicates that the node was having problems warm starting. The rest of the log would provide more context about the cause of the restart.

Thank you for your time,

-DM