1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
|
use std::time::Duration;
use garage_util::data::*;
use garage_util::error::OkOrMessage;
use garage_util::time::*;
use garage_table::util::*;
use crate::bucket_table::*;
use crate::garage::Garage;
use crate::helper::error::*;
use crate::key_table::*;
use crate::s3::object_table::*;
pub struct BucketHelper<'a>(pub(crate) &'a Garage);
#[allow(clippy::ptr_arg)]
impl<'a> BucketHelper<'a> {
pub async fn resolve_global_bucket_name(
&self,
bucket_name: &String,
) -> Result<Option<Uuid>, Error> {
// Bucket names in Garage are aliases, true bucket identifiers
// are 32-byte UUIDs. This function resolves bucket names into
// their full identifier by looking up in the bucket_alias_table.
// This function also allows buckets to be identified by their
// full UUID (hex-encoded). Here, if the name to be resolved is a
// hex string of the correct length, it is directly parsed as a bucket
// identifier which is returned. There is no risk of this conflicting
// with an actual bucket name: bucket names are max 63 chars long by
// the AWS spec, and hex-encoded UUIDs are 64 chars long.
let hexbucket = hex::decode(bucket_name.as_str())
.ok()
.and_then(|by| Uuid::try_from(&by));
if let Some(bucket_id) = hexbucket {
Ok(self
.0
.bucket_table
.get(&EmptyKey, &bucket_id)
.await?
.filter(|x| !x.state.is_deleted())
.map(|_| bucket_id))
} else {
Ok(self
.0
.bucket_alias_table
.get(&EmptyKey, bucket_name)
.await?
.and_then(|x| *x.state.get()))
}
}
#[allow(clippy::ptr_arg)]
pub async fn resolve_bucket(&self, bucket_name: &String, api_key: &Key) -> Result<Uuid, Error> {
let api_key_params = api_key
.state
.as_option()
.ok_or_message("Key should not be deleted at this point")?;
if let Some(Some(bucket_id)) = api_key_params.local_aliases.get(bucket_name) {
Ok(*bucket_id)
} else {
Ok(self
.resolve_global_bucket_name(bucket_name)
.await?
.ok_or_else(|| Error::NoSuchBucket(bucket_name.to_string()))?)
}
}
/// Returns a Bucket if it is present in bucket table,
/// even if it is in deleted state. Querying a non-existing
/// bucket ID returns an internal error.
pub async fn get_internal_bucket(&self, bucket_id: Uuid) -> Result<Bucket, Error> {
Ok(self
.0
.bucket_table
.get(&EmptyKey, &bucket_id)
.await?
.ok_or_message(format!("Bucket {:?} does not exist", bucket_id))?)
}
/// Returns a Bucket if it is present in bucket table,
/// only if it is in non-deleted state.
/// Querying a non-existing bucket ID or a deleted bucket
/// returns a bad request error.
pub async fn get_existing_bucket(&self, bucket_id: Uuid) -> Result<Bucket, Error> {
self.0
.bucket_table
.get(&EmptyKey, &bucket_id)
.await?
.filter(|b| !b.is_deleted())
.ok_or_else(|| Error::NoSuchBucket(hex::encode(bucket_id)))
}
// ----
pub async fn is_bucket_empty(&self, bucket_id: Uuid) -> Result<bool, Error> {
let objects = self
.0
.object_table
.get_range(
&bucket_id,
None,
Some(ObjectFilter::IsData),
10,
EnumerationOrder::Forward,
)
.await?;
if !objects.is_empty() {
return Ok(false);
}
#[cfg(feature = "k2v")]
{
use garage_rpc::ring::Ring;
use std::sync::Arc;
let ring: Arc<Ring> = self.0.system.ring.borrow().clone();
let k2vindexes = self
.0
.k2v
.counter_table
.table
.get_range(
&bucket_id,
None,
Some((DeletedFilter::NotDeleted, ring.layout.node_id_vec.clone())),
10,
EnumerationOrder::Forward,
)
.await?;
if !k2vindexes.is_empty() {
return Ok(false);
}
}
Ok(true)
}
// ----
/// Deletes all incomplete multipart uploads that are older than a certain time.
/// Returns the number of uploads aborted.
/// This will also include non-multipart uploads, which may be lingering
/// after a node crash
pub async fn cleanup_incomplete_uploads(
&self,
bucket_id: &Uuid,
older_than: Duration,
) -> Result<usize, Error> {
let older_than = now_msec() - older_than.as_millis() as u64;
let mut ret = 0usize;
let mut start = None;
loop {
let objects = self
.0
.object_table
.get_range(
bucket_id,
start,
Some(ObjectFilter::IsUploading {
check_multipart: None,
}),
1000,
EnumerationOrder::Forward,
)
.await?;
let abortions = objects
.iter()
.filter_map(|object| {
let aborted_versions = object
.versions()
.iter()
.filter(|v| v.is_uploading(None) && v.timestamp < older_than)
.map(|v| ObjectVersion {
state: ObjectVersionState::Aborted,
uuid: v.uuid,
timestamp: v.timestamp,
})
.collect::<Vec<_>>();
if !aborted_versions.is_empty() {
Some(Object::new(
object.bucket_id,
object.key.clone(),
aborted_versions,
))
} else {
None
}
})
.collect::<Vec<_>>();
ret += abortions.len();
self.0.object_table.insert_many(abortions).await?;
if objects.len() < 1000 {
break;
} else {
start = Some(objects.last().unwrap().key.clone());
}
}
Ok(ret)
}
}
|
