Technically speaking, the current code is not wrong. However, it was written
before we had do_with, and I ended up dowithing it while chasing our erratic
bug under the suspicion that this code could somehow be related with our bug.
Turns out it isn't, but now that I went through the trouble of dowithing it -
and since do_with is easier to reason about and guarantee liveness, let's go
with this option.
Signed-off-by: Glauber Costa <glommer@cloudius-systems.com>
Reviewed-by: Pekka Enberg <penberg@cloudius-systems.com>
When a promise that still tracks its future is destroyed, but after
future::get() or future::then() was called, the promise thinks it was
destroyed before generating a value, and fails an assertion.
Fix be detaching the promise from the future as soon as get() or then()
remove the value.
cql_query_test segfaults in debug mode due to infinite recursion (see
trace below). The problem started to appear after Avi's always-defer
change.
It looks like the following boost::any() constructor template is
instantiated when boost::any is copied into a lambda:
// Perfect forwarding of ValueType
template<typename ValueType>
any(ValueType&& value, typename boost::disable_if<boost::is_same<any&, ValueType> >::type* = 0)
: content(new holder< typename remove_reference<ValueType>::type >(static_cast<ValueType&&>(value)))
{
}
It checks for any& to disable itself on forwarding, but it doesn't
check for const any&. There are (non-template) constructors which take
const any&, but I guess some C++ rule favors the template
version. This results in infinite recursion of constructor invocations
between any() and holder().
Workaround is to make the lambda mutable, so that the argument type is
any& and template doesn't kick in.
Trace:
CU 0x4debed5, DIE 0x4f176e6>, this=0x602000044550) at /usr/include/boost/any.hpp:175
ql_query_test, CU 0x4debed5, DIE 0x4f6c509>, this=0x602000044518) at /usr/include/boost/any.hpp:72
CU 0x4debed5, DIE 0x4f176e6>, this=0x602000044510) at /usr/include/boost/any.hpp:175
ql_query_test, CU 0x4debed5, DIE 0x4f6c509>, this=0x6020000444e8) at /usr/include/boost/any.hpp:72
CU 0x4debed5, DIE 0x4f176e6>, this=0x6020000444e0) at /usr/include/boost/any.hpp:175
oost::any const>, void>::type*) (this=0x7ffff2ffa6f0, value=<unknown type in /home/tgrabiec/src/urchin2/build
/release/tests/urchin/cql_query_test, CU 0x4debed5, DIE 0x4f6c509>) at /usr/include/boost/any.hpp:72
5u> const&, std::vector<boost::any, std::allocator<boost::any> >, std::vector<boost::any, std::allocator<boos
t::any> >, basic_sstring<char, unsigned int, 15u> const&, boost::any)::{lambda(database&)#1}::operator()(data
base&) const::{lambda(std::unique_ptr<mutation_partition const, std::default_delete<mutation_partition> >)#1}
::unique_ptr(std::unique_ptr<mutation_partition const, std::default_delete<mutation_partition> >&&) (this=0x7
ffff2ffa6c0) at tests/urchin/cql_test_env.cc:126
We want query_local() to actually respect the key we pass to it. Fixes
an issue in keyspace merging code where we returned multiple rows for a
keyspace.
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>
From Pekka:
"This series fixes two issues in schema merging code that caused
cql_query_test to sometimes think that two identical keyspaces are
different. This forced the code to use the alter keyspace paths which
are not implemented and just throw an exception."
We must compare types with ->equal(). Fixes keyspace merging issues
where the merging code mistakenly thinks two keyspaces are different and
calls the alter keyspace path which is not implemented.
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>
Fix the check in read_schema_for_keyspaces() to not insert empty
result sets in the return value. There's no functional change as the
merge algorithms already deal with the case. However, the code is now
closer to what origin does.
Spotted while reading the code.
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>
This reverts commit a19d2171eb.
This commit breaks cql_query_test.
[asias@hjpc urchin]$ ./cql_query_test
Running 1 test case...
WARNING: Not implemented: COMPACT_TABLES
WARNING: Not implemented: METRICS
WARNING: Not implemented: PERMISSIONS
cql_query_test: core/distributed.hh:290: Service&
distributed<Service>::local() [with Service =
service::storage_service]: Assertion `local_is_initialized()' failed.
unknown location(0): fatal error in "test_create_keyspace_statement":
signal: SIGABRT (application abort requested)
tests/test-utils.cc(31): last checkpoint
*** 1 failure detected in test suite "tests/urchin/cql_query_test.cc"
(gdb) bt
#0 0x00000032930348d7 in __GI_raise (sig=sig@entry=6) at
../sysdeps/unix/sysv/linux/raise.c:55
#1 0x000000329303653a in __GI_abort () at abort.c:89
#2 0x000000329302d47d in __assert_fail_base (fmt=0x3293186cb8
"%s%s%s:%u: %s%sAssertion `%s' failed.\n%n",
assertion=assertion@entry=0x8ec10a "local_is_initialized()",
file=file@entry=0x92508d "core/distributed.hh",
line=line@entry=290, function=function@entry=0x8ed440
<distributed<service::storage_service>::local()::__PRETTY_FUNCTION__>
"Service& distributed<Service>::local() [with Service =
service::storage_service]")
at assert.c:92
#3 0x000000329302d532 in __GI___assert_fail (assertion=0x8ec10a
"local_is_initialized()", file=0x92508d "core/distributed.hh",
line=290,
function=0x8ed440
<distributed<service::storage_service>::local()::__PRETTY_FUNCTION__>
"Service& distributed<Service>::local() [with Service =
service::storage_service]") at assert.c:101
#4 0x0000000000430f19 in local (this=<optimized out>) at
core/distributed.hh:290
#5 get_local_storage_service () at service/storage_service.hh:3326
#6 keyspace::create_replication_strategy (this=0x7ffff6bf8350) at
database.cc:690
#7 0x000000000061537a in
_ZZZN2db20legacy_schema_tables15merge_keyspacesERN7service13storage_proxyEOSt3mapI13basic_sstringIcjLj15EE13lw_shared_ptrIN5query10result_setEESt4lessIS6_ESaISt4pairIKS6_SA_EEESI_ENKUlRT_E0_clISt6ve
ctorISF_SG_EEEDaSK_ENKUlR8databaseE_clESQ_ () at
db/legacy_schema_tables.cc:584
#8 0x0000000000617d19 in operator() (__closure=0x7ffff6bf8650) at
./core/distributed.hh:284
In the test, storage_service and other services are not stared.
Let's revert it and figure out a way to run cql_query_test with the
needed services started properly and then bring the "storage_service:
Remove ad-hoc token_metadata creation" change back.
Following std::async(), seastar::async(func) causes func to be executed
in a seastar thread, where it can block using future<>::get(). Whatever
func returns is converted to a future, and returned as async()s return
value.
Sometimes remote data has to be copied to local cpu, but if data is
already local copy can be avoided. Introduce helper function that moves
or copies data depending on origin cpu.
Optimize checksum calculation by deferring "full checksum" update until
we've computed a full per-chunk checksum.
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>
Code calls failure_detector::is_alive on all cpus, so we start
failure_detector on all cpus. However, the internal data of failure_detector
is modified on cpu zero and it is not replicated to non-zero cpus.
This is fine since the user of failure_detector (the gossiper) accesses
it on cpu0 only.
_token_metadata is needed by replication strategy code on all cpus.
Changes to _token_metadata are done on cpu 0. Replicate it to all cpus.
We may copy only if _token_metadata actually changes. As a starter, we
always copy in gossip modification callbacks.
We do not care about the order of the tokens.
Also, in token_metadata, we use unordered_set for tokens as well, e.g.
update_normal_tokens. Unify the usage.
