+// Copyright (C) The Arvados Authors. All rights reserved.
+//
+// SPDX-License-Identifier: AGPL-3.0
+
package main
import (
var page arvados.CollectionList
var zero int
params.Limit = &zero
+ params.Count = "exact"
err := c.RequestAndDecode(&page, "GET", "arvados/v1/collections", nil, params)
return page.ItemsAvailable, err
}
}
expectCount, err := countCollections(c, arvados.ResourceListParams{
- IncludeTrash: true,
+ IncludeTrash: true,
+ IncludeOldVersions: true,
})
if err != nil {
return err
}
+ // Note the obvious way to get all collections (sorting by
+ // UUID) would be much easier, but would lose data: If a
+ // client were to move files from collection with uuid="zzz"
+ // to a collection with uuid="aaa" around the time when we
+ // were fetching the "mmm" page, we would never see those
+ // files' block IDs at all -- even if the client is careful to
+ // save "aaa" before saving "zzz".
+ //
+ // Instead, we get pages in modified_at order. Collections
+ // that are modified during the run will be re-fetched in a
+ // subsequent page.
+
limit := pageSize
if limit <= 0 {
// Use the maximum page size the server allows
limit = 1<<31 - 1
}
params := arvados.ResourceListParams{
- Limit: &limit,
- Order: "modified_at, uuid",
- Select: []string{"uuid", "manifest_text", "modified_at", "portable_data_hash", "replication_desired"},
- IncludeTrash: true,
+ Limit: &limit,
+ Order: "modified_at, uuid",
+ Count: "none",
+ Select: []string{"uuid", "unsigned_manifest_text", "modified_at", "portable_data_hash", "replication_desired"},
+ IncludeTrash: true,
+ IncludeOldVersions: true,
}
var last arvados.Collection
var filterTime time.Time
callCount := 0
+ gettingExactTimestamp := false
for {
progress(callCount, expectCount)
var page arvados.CollectionList
}
last = coll
}
- if last.ModifiedAt == nil || *last.ModifiedAt == filterTime {
- if page.ItemsAvailable > len(page.Items) {
- // TODO: use "mtime=X && UUID>Y"
- // filters to get all collections with
- // this timestamp, then use "mtime>X"
- // to get the next timestamp.
- return fmt.Errorf("BUG: Received an entire page with the same modified_at timestamp (%v), cannot make progress", filterTime)
- }
+ if len(page.Items) == 0 && !gettingExactTimestamp {
break
+ } else if last.ModifiedAt == nil {
+ return fmt.Errorf("BUG: Last collection on the page (%s) has no modified_at timestamp; cannot make progress", last.UUID)
+ } else if len(page.Items) > 0 && *last.ModifiedAt == filterTime {
+ // If we requested time>=X and never got a
+ // time>X then we might not have received all
+ // items with time==X yet. Switch to
+ // gettingExactTimestamp mode (if we're not
+ // there already), advancing our UUID
+ // threshold with each request, until we get
+ // an empty page.
+ gettingExactTimestamp = true
+ params.Filters = []arvados.Filter{{
+ Attr: "modified_at",
+ Operator: "=",
+ Operand: filterTime,
+ }, {
+ Attr: "uuid",
+ Operator: ">",
+ Operand: last.UUID,
+ }}
+ } else if gettingExactTimestamp {
+ // This must be an empty page (in this mode,
+ // an unequal timestamp is impossible) so we
+ // can start getting pages of newer
+ // collections.
+ gettingExactTimestamp = false
+ params.Filters = []arvados.Filter{{
+ Attr: "modified_at",
+ Operator: ">",
+ Operand: filterTime,
+ }}
+ } else {
+ // In the normal case, we know we have seen
+ // all collections with modtime<filterTime,
+ // but we might not have seen all that have
+ // modtime=filterTime. Hence we use >= instead
+ // of > and skip the obvious overlapping item,
+ // i.e., the last item on the previous
+ // page. In some edge cases this can return
+ // collections we have already seen, but
+ // avoiding that would add overhead in the
+ // overwhelmingly common cases, so we don't
+ // bother.
+ filterTime = *last.ModifiedAt
+ params.Filters = []arvados.Filter{{
+ Attr: "modified_at",
+ Operator: ">=",
+ Operand: filterTime,
+ }, {
+ Attr: "uuid",
+ Operator: "!=",
+ Operand: last.UUID,
+ }}
}
- filterTime = *last.ModifiedAt
- params.Filters = []arvados.Filter{{
- Attr: "modified_at",
- Operator: ">=",
- Operand: filterTime,
- }, {
- Attr: "uuid",
- Operator: "!=",
- Operand: last.UUID,
- }}
}
progress(callCount, expectCount)
Attr: "modified_at",
Operator: "<=",
Operand: filterTime}},
- IncludeTrash: true,
+ IncludeTrash: true,
+ IncludeOldVersions: true,
}); err != nil {
return err
} else if callCount < checkCount {