fee-mendes
fee-mendes
It turns out that an user can rightfully have a aio-nr value bigger than aio-max-nr. Such situation happens when aio-max-nr value is modified on the fly and the running system...
Within has_enough_aio_nr(), the check has the following logic: aio_max_nr - aio_nr < reactor::max_aio * smp::count which one would naively assume is correct. However, as defined within reactor.cc, smp::count evaluates to...
It turns out that an user can rightfully have a `aio-nr` value bigger than `aio-max-nr`. Such situation happens when `aio-max-nr` value is modified on the fly and the running system...
In a situation when the running system concurrent AIO contexts is exhausted (eg: `aio-max-nr == aio-nr`), seastar will fail to fallback to epoll backend. This happens for two reasons: `reactor_backend_selector::available`...
perftune's [__check_host_type](https://github.com/scylladb/seastar/blob/master/scripts/perftune.py#L458:L477) method uses IMDSv1 for retrieving instance metadata. It turns out that - as described in https://github.com/scylladb/scylla/issues/10490 - Ec2 instances may have only [IMDSv2](https://aws.amazon.com/blogs/security/defense-in-depth-open-firewalls-reverse-proxies-ssrf-vulnerabilities-ec2-instance-metadata-service/) HTTP calls allowed. When requests...
Scylla version (or git commit hash): `5.0.1-0.20220719.b177dacd3` with build-id `217f31634f8c8722cadcfe57ade8da58af05d415` Scylla Alternator documentation properly states that [the BillingMode and ProvisionedThroughput options are ignored by Scylla](https://github.com/scylladb/scylladb/blob/master/docs/alternator/compatibility.md#provisioning). It is not clear -...
**Describe the bug** `scyllaArgs` value is not respected when deploying via Helm as it is not defined within [helm/scylla/templates/scyllacluster.yaml](https://github.com/scylladb/scylla-operator/blob/master/helm/scylla/templates/scyllacluster.yaml). Notice this is only for Helm deployments, it works just fine...
The default Scylla Manager behavior is to detect whether a cluster supports `RowLevelRepair` and - if it does - default to it. Otherwise, we fallback to the old default controller...
Consider an overprovisioned cluster with 9 nodes a 1 single user keyspace to backup. In particular, the data set is small enough that it can fit in memory, but large...
As of commits https://github.com/scylladb/java-driver/commit/d184387961cc439b2c9379d90868fe7b75ad2284 and https://github.com/scylladb/java-driver/commit/d2c151331a1d1fd29c12f678f16bb66974e8aa0d, it would be nice to cut off a driver release with such capabilities. We have users asking for it and, nonetheless, these commits are...