Each outgoing_file_message might contain multiple mutations. Send them
one mutation per RPC call (using frozen_mutation), instead of one big
outgoing_file_message per one RPC call.
It is common for some operations, like system table updates, to try and guarantee
some particular ordering of operations.
The way Origin does it, is by simply incrementing one to the current timestamp.
Our calls, however, are being dispatched through our internal query processor, which
has a builtin client_state.
Our client_state has a mechanism to guarantee monotonicity, by adding 1 if needed
to operations that happen subsequentially. By using a clock that is not wired up
to this mechanism, we can't really guarantee that if other operations happened to
get in between.
If we expose this mechanism through the query_processor, we will be able to guarantee
that.
Signed-off-by: Glauber Costa <glommer@cloudius-systems.com>
Fix release version to be compatible with origin to avoid confusing
clients.
Before:
[penberg@nero apache-cassandra-2.1.7]$ ./bin/cqlsh --no-color 127.0.0.1
Connected to Test Cluster at 127.0.0.1:9042.
[cqlsh 5.0.1 | Cassandra SeastarDB v0.1 | CQL spec 3.2.0 | Native protocol v3]
Use HELP for help.
cqlsh>
After:
[penberg@nero apache-cassandra-2.1.7]$ ./bin/cqlsh --no-color 127.0.0.1
Connected to Test Cluster at 127.0.0.1:9042.
[cqlsh 5.0.1 | Cassandra 2.2.0 | CQL spec 3.2.0 | Native protocol v3]
Use HELP for help.
cqlsh>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>
"As previously said, there were some unidentified bugs that prevented update_tokens from
working properly. The first one was sent alongside the series, the second one took me more
time, but it is fixed here."
"(Partial?) implementation of BatchLogManager.
Requires the token function/restriction series.
Functional as in that it can create batchlog mutations, and do replay
of data in this system table.
Since range queries does not yet work, it only handles a very small
table contents.
It is not used yet either, but will eventually be needed for batch statements
etc."
"This patches prevent compression validation code from rejecting statements
with an empty sstable_compression property. Correct behaviour in such cases
is not to use compression at all.
This would have been a 10 minuts fix if antlr3 didn't prepare surprise of its
own. Empty STRING_LITERALS weren't handled properly and a workaround for that
problem was introduced."
Somewhat simplifies version of the Origin code, since from what I
can see, there is less need for us to do explicit query sends in
the BLM itself, instead we can just go through storage_proxy.
I could be wrong though.
The serialization for "already exists" error is special as explained in
the CQL binary protocol specification:
0x2400 Already_exists: The query attempted to create a keyspace or a
table that was already existing. The rest of the ERROR message
body will be <ks><table> where:
<ks> is a [string] representing either the keyspace that
already exists, or the keyspace in which the table that
already exists is.
<table> is a [string] representing the name of the table that
already exists. If the query was attempting to create a
keyspace, <table> will be present but will be the empty
string.
Fix that to unconfuse "cqlsh" when attemting to create a duplicate
keyspace or table.
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>
setText() is meant to override the token value. However, when antlr3 is
asked to return the token value and needs to determine whether setText()
was called it checks if the string set by setText() is empty or not.
This basically means that it is impossible to override token value with
an empty string.
This problem is solved by using a string with a single byte set to -1 to
represent empty string.
Signed-off-by: Paweł Dziepak <pdziepak@cloudius-systems.com>
"Implementation of:
* Token func
* Token restriction
* Token relation
* Token cql parsing
This series contains some hefty refactoring of the cql3::restrictions
interfaces, to handle the slightly different conditions of the token
relation. (See individual commit comment)"
Needed to reasonably cleanly implement token restrictions.
* Fixed constness for various virtuals.
* primary_key_restrictions now inherit abstract_restriction,
similuar to Origin (for better or for worse), to avoid
duplicating attributes etc.
* primary_key_restrictions bounds & values renamed (so not to
collide with restriction), and some logic pushed downwards
(building bounds), to avoid abstraction breakage in
statement_restrictions
* primary_key_restrictions merging is now potentially replacing
to make dispatching token/multicolumn restrictions simpler
