Routing log-based information
Routing log-based information between production servers and logging servers is disclosed. A log entry for a logging server is generated at a production server. A shard identifier is computed for a shard associated with the logging server based on application of a hashing algorithm to properties associated with the production server. The hashing algorithm and properties are selected to prevent or minimize the likelihood of computing of the same shard identifier by another production server for the same shard associated with the logging server. The log entry is transmitted to the shard associated with the logging server. A determination is made that the logging server has malfunctioned by detecting that the log entry transmitted to the shard is absent. In response, another shard identifier is computed for another shard of another logging server and a subsequent log entry from the production server is transmitted to the another shard of the another logging over. No load balancers are used by the routing system.
1. A computer-implemented method, comprising:
generating a log entry for a logging server at a production server;
computing a shard identifier for a shard associated with the logging server, wherein
the shard identifier is computed based on application of a hashing algorithm to one or more properties associated with the production server, and
the hashing algorithm and the one or more properties are selected to prevent computation of the shard identifier by another production server for the shard associated with the logging server;
transmitting the log entry to the shard associated with the logging server;
determining that the logging server has malfunctioned by detecting that the log entry transmitted to the shard is absent;
in response to detecting failure of the logging server, computing another shard identifier for another shard of another logging server;
transmitting a subsequent log entry from the production server to the another shard of the another logging server; and
removing the logging server that has malfunctioned by changing a modulo value used in the computation of the shard identifier.
2. The computer-implemented method of claim 1 , wherein
the shard identifier is static, and
the shard identifier is mapped one-to-one onto the shard associated with the logging server.
3. The computer-implemented method of claim 1 , wherein
a property of the one or more properties associated with the production server is an instance identifier associated with a virtual machine implementing the production server.
4. The computer-implemented method of claim 1 , further comprising:
determining a time at which the subsequent log entry was routed from the shard of the logging server to the another shard of the another logging server.
5. The computer-implemented method of claim 1 , wherein
the hashing algorithm comprises CRC32, MD5, or SHA256.
6. The computer-implemented method of claim 4 , wherein
the hashing algorithm and the one or more properties are selected to reduce a likelihood the another production server computes the shard identifier for the shard associated with the logging server instead of the another shard identifier for the another shard associated with the another logging server.
7. A non-transitory computer readable storage medium comprising program instructions executable to:
generate a log entry for a logging server at a production server;
compute a shard identifier for a shard associated with the logging server, wherein
the shard identifier is computed based on application of a hashing algorithm to one or more properties associated with the production server, and
the hashing algorithm and the one or more properties are selected to prevent computation of the shard identifier by another production server for the shard associated with the logging server;
transmit the log entry to the shard associated with the logging server;
determine that the logging server has malfunctioned by detecting that the log entry transmitted to the shard is absent;
in response to detecting failure of the logging server, compute another shard identifier for another shard of another logging server;
transmit a subsequent log entry from the production server to the another shard of the another logging server; and
remove the logging server that has malfunctioned by changing a modulo value used in the computation of the shard identifier.
8. The non-transitory computer readable storage medium of claim 7 , wherein
the shard identifier is static, and
the shard identifier is mapped one-to-one onto the shard associated with the logging server.
9. The non-transitory computer readable storage medium of claim 7 , wherein
a property of the one or more properties associated with the production server is an instance identifier associated with a virtual machine implementing the production server.
10. The non-transitory computer readable storage medium of claim 7 , further comprising:
determining a time at which the subsequent log entry was routed from the shard of the logging server to the another shard of the another logging server.
11. The non-transitory computer readable storage medium of claim 7 , wherein
the hashing algorithm comprises CRC32, MD5, or SHA256.
12. The non-transitory computer readable storage medium of claim 10 , wherein
the hashing algorithm and the one or more properties are selected to reduce a likelihood the another production server computes the shard identifier for the shard associated with the logging server instead of the another shard identifier for the another shard associated with the another logging server.
13. A system comprising:
one or more processors; and
a memory coupled to the one or more processors, wherein the memory stores program instructions executable by the one or more processors to:
generate a log entry for a logging server at a production server;
compute a shard identifier for a shard associated with the logging server, wherein
the shard identifier is computed based on application of a hashing algorithm to one or more properties associated with the production server, and
the hashing algorithm and the one or more properties are selected to prevent computation of the shard identifier by another production server for the shard associated with the logging server;
transmit the log entry to the shard associated with the logging server;
determine that the logging server has malfunctioned by detecting that the log entry transmitted to the shard is absent;
in response to detecting failure of the logging server, compute another shard identifier for another shard of another logging server;
transmit a subsequent log entry from the production server to the another shard of the another logging server; and
remove the logging server that has malfunctioned by changing a modulo value used in the computation of the shard identifier.
14. The system of claim 13 , wherein
the shard identifier is static, and
the shard identifier is mapped one-to-one onto the shard associated with the logging server.
15. The system of claim 13 , wherein
a property of the one or more properties associated with the production server is an instance identifier associated with a virtual machine implementing the production server.
16. The system of claim 13 , further comprising:
determining a time at which the subsequent log entry was routed from the shard of the logging server to the another shard of the another logging server.
17. The system of claim 13 , wherein
the hashing algorithm comprises CRC32, MD5, or SHA256.