Commit Graph

21 Commits

Author SHA1 Message Date
Pavel Emelyanov
ae28814b1c token: Introduce raw() helper and raw comparator
In next patches the entries having token on-board will be
moved onto B+-tree rails. For this the int64_t value of the
token will be used as B+ key, so prepare for this.

One corner case -- the after_all_keys tokens must be resolved
to int64::max value to appear at the "end" of the tree. This
is not the same as "before_all_keys" case, which maps to the
int64::min value which is not allowed for regular tokens. But
for the sake of B+ switch this is OK, the conflicts of token
raw values are explicitly resolved in next patches.

Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
2020-07-14 16:30:02 +03:00
Pavel Emelyanov
6d7ae4ead1 dht: Mark noexcept methods
These are either trivially noexcept already, or call each-other, thus becoming noexcept too

Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
2020-07-09 14:41:03 +03:00
Piotr Jastrzebski
1d1c6af72a dht: Add find_first_token_for_shard
This new function finds the first token in range (start, end] that
belongs to given shard.

Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-04-22 18:24:54 +02:00
Piotr Jastrzebski
c82adb7906 dht: use long_token in token::to_int64
Previous implementation of to_int64 wasn't handling dht::minimum_token
and dht::maximum_token.

Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-04-22 16:12:00 +02:00
Piotr Jastrzebski
2aaf81bf7c dht: Exclude -2^63 value from get_random_token
-2^63 is a value reserved for min/max token boundaries and shouldn't be used for
regular tokens. This patch fixes get_random_token to never create token with
value -2^63. On the way dht::get_random_number template method is removed
because it was exclusively used by get_random_token.

Also use uniform_int_distribution with int64_t instead of uint64_t by using
correct constructor parameter that guarantees values between -2^63+1 and 2^63-1
inclusively.

Tests: unit(dev)

Fixes #6237.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Message-Id: <0a1a939355f5005039d5c2c7c513bad94cf60be2.1587302093.git.piotr@scylladb.com>
2020-04-19 18:17:35 +03:00
Rafael Ávila de Espíndola
f3fd466156 dht: Use get_random_number<uint64_t> instead of int64_t in token::get_random_token
I bisect the opposite change in
9c202b52da as the cause of issue 6193. I
don't know why. Maybe get_random_number<signed_type> is buggy?

In any case, reverting to uint64_t solves the issue.

Fixes #6193

Signed-off-by: Rafael Ávila de Espíndola <espindola@scylladb.com>
Message-Id: <20200418001611.440733-1-espindola@scylladb.com>
2020-04-19 09:46:06 +03:00
Piotr Jastrzebski
968177da04 cdc: store tokens in cdc description as longs
Previously the tokens were stored as strings
because token could have been represented in multiple ways.
Now token representation is always int64_t so we can
store them as ints in cdc description as well.

Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-03-06 11:59:59 +01:00
Piotr Jastrzebski
6d24f26ff7 token: cast to int64_t not long in long_token
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
50cfe81331 murmur3: move sharding logic to token and i_partitioner
Since token representation is fixed now, all the partitioners
will share the sharding logic. It makes sense now to keep
the logic in common super class and separate header that's
included only in i_partitioner.cc.

shard_of and token_for_next_shard are now implemented in
i_partitioner. They would be non-virtual but we have to
keep them virtual because one test is overriding them
to enforce some specific sharding.

Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
d4d55160f0 partitioner: move get_token_validator to token
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
2c630c5820 partitioner: merge tri_compare into dht::tri_compare
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
d0d8bfaf8c partitioner: move describe_ownership to token
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
f845220445 partitioner: move from_bytes to token
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
8107d99e3d partitioner: move from_string to token
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
03bdce2d68 partitioner: move to_sstring to token
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
9c202b52da partitioner: move get_random_token to token
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
f42b1ee819 partitioner: move midpoint function to token
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
1d1ac476c3 token: remove token_view
Now that both token and token_view contain int64_t
it makes no sense to keep the view.

Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
05e0451b27 token: change _data to int64_t
Previously _data was stored as array of 8 bytes in
network byte order.
After this change it stores the same value in int64_t
in host byte order.

Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:31:32 +01:00
Piotr Jastrzebski
b569d127a0 token: change data to array<uint8_t, 8>
It is save to do such change because we support only
Murmur3Partitioner which uses only tokens that are
8 bytes long.

Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:30:46 +01:00
Piotr Jastrzebski
0da21c28ab partitioner: Extract token to separate .hh and .cc files
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
2020-02-05 09:18:24 +01:00